firefrost-gaming/skill-seekers-reference

Files

yusyus 6fded977dd feat: add Kotlin language support for codebase analysis (#287 )

Adds full C3.x pipeline support for Kotlin (.kt, .kts):
- Language detection patterns (40+ weighted patterns for data/sealed classes, coroutines, companion objects, KMP, etc.)
- AST regex parser in code_analyzer.py (classes, objects, functions, extension functions, suspend functions)
- Dependency extraction for Kotlin import statements (with alias support)
- Design pattern adaptations (object→Singleton, companion→Factory, sealed→Strategy, data→Builder, Flow→Observer)
- Test example extraction for JUnit 4/5, Kotest, MockK, Spek
- Config detection for build.gradle.kts / settings.gradle.kts
- Extension maps registered in codebase_scraper, unified_codebase_analyzer, github_scraper, generate_router

Also fixes pre-existing parser count tests (35→36 for doctor command added in previous commit).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-03-28 23:25:12 +03:00

Changelog

All notable changes to Skill Seeker will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

Added

Kotlin language support for codebase analysis — Full C3.x pipeline support: AST parsing (classes, objects, functions, data/sealed classes, extension functions, coroutines), dependency extraction, design pattern recognition (object declaration→Singleton, companion object→Factory, sealed class→Strategy), test example extraction (JUnit, Kotest, MockK, Spek), language detection patterns, config detection (build.gradle.kts), and extension maps across all analyzers (#287)
Headless browser rendering (--browser flag) — uses Playwright to render JavaScript SPA sites (React, Vue, etc.) that return empty HTML shells. Auto-installs Chromium on first use. Optional dep: pip install "skill-seekers[browser]" (#321)
skill-seekers doctor command — 8 diagnostic checks (Python version, package install, git, core/optional deps, API keys, MCP server, output dir) with pass/warn/fail status and --verbose flag (#316)
Prompt injection check workflow — bundled prompt-injection-check workflow scans scraped content for injection patterns (role assumption, instruction overrides, delimiter injection, hidden instructions). Added as first stage in default and security-focus workflows. Flags suspicious content without removing it (#324)
6 behavioral UML diagrams — 3 sequence (create pipeline, GitHub+C3.x flow, MCP invocation), 2 activity (source detection, enhancement pipeline), 1 component (runtime dependencies with interface contracts)

Fixed

GitHub language detection crashes with TypeError when API response contains non-integer metadata keys (e.g., "url") — now filters to integer values only (#322)
C3.x codebase analysis crashes with TypeError — _run_c3_analysis() and _analyze_c3x() passed removed enhance_with_ai/ai_mode kwargs to analyze_codebase() instead of enhance_level (#323)

[3.4.0] - 2026-03-21

Added

OpenCode adaptor (--target opencode) - Directory-based packaging with dual-format YAML frontmatter
OpenAI-compatible base class - Shared base for all OpenAI-compatible LLM platforms
6 new LLM platform adaptors: Kimi (--target kimi), DeepSeek (--target deepseek), Qwen (--target qwen), OpenRouter (--target openrouter), Together AI (--target together), Fireworks AI (--target fireworks)
7 new CLI agent install paths: roo, cline, aider, bolt, kilo, continue, kimi-code (total: 18 agents)
OpenCode skill splitter - Auto-split large docs into focused sub-skills with router
Bi-directional skill converter - Import/export between OpenCode and any platform format
Distribution files for Smithery (smithery.yaml), GitHub Actions (templates/github-actions/update-skills.yml), and Claude Code Plugin
Full UML architecture documentation — 14 class diagrams synced from source code via StarUML
StarUML HTML API reference documentation export
Ecosystem section in README linking all Skill Seekers repos (PyPI, website, plugin, GitHub Action)

Fixed

sanitize_url() crashes on Python 3.14 due to strict urlparse rejecting bracket-containing URLs (#284)
Blindly appending /index.html.md to non-.md URLs — now only appends for URLs that should have it (#277)
Unified scraper temp config uses unified format for doc_scraper instead of raw args (#317)
Unicode arrows in CLI help text replaced with ASCII for Windows cp1252 compatibility
CLI flags in plugin slash commands corrected (create uses --preset, package uses --target)
MiniMax adaptor improvements from PR #318 review (#319)
Misleading "Scraped N pages" count reported visited URLs instead of saved pages — now shows (N saved, M skipped) (#320)
"No scraped data found" after successful scrape on JavaScript SPA sites — now warns that site requires JS rendering (#320, #321)

Changed

Refactored MiniMax adaptor to inherit from shared OpenAI-compatible base class
Platform count: 5 → 12 LLM targets
Agent count: 11 → 18 install paths
Consolidated Docs/ into docs/ (single documentation directory)
Removed stale root-level test scripts and junk files
Removed stale UNIFIED_PARSERS.md superseded by UML architecture
Added architecture references to README.md and CONTRIBUTING.md
Fixed pre-existing ruff format issues in 5 files

[3.3.0] - 2026-03-16

Theme: 10 new source types (17 total), EPUB unified integration, sync-config command, performance optimizations, 12 README translations, and 19 bug fixes. 117 files changed, +41,588 lines since v3.2.0.

Supported Source Types (17)

#	Type	CLI Command	Config Type	Auto-Detection
1	Documentation (web)	`scrape` / `create <url>`	`documentation`	HTTP/HTTPS URLs
2	GitHub repository	`github` / `create owner/repo`	`github`	`owner/repo` or github.com URLs
3	PDF document	`pdf` / `create file.pdf`	`pdf`	`.pdf` extension
4	Word document	`word` / `create file.docx`	`word`	`.docx` extension
5	EPUB e-book	`epub` / `create file.epub`	`epub`	`.epub` extension
6	Video	`video` / `create <url/file>`	`video`	YouTube/Vimeo URLs, video extensions
7	Local codebase	`analyze` / `create ./path`	`local`	Directory paths
8	Jupyter Notebook	`jupyter` / `create file.ipynb`	`jupyter`	`.ipynb` extension
9	Local HTML	`html` / `create file.html`	`html`	`.html`/`.htm` extensions
10	OpenAPI/Swagger	`openapi` / `create spec.yaml`	`openapi`	`.yaml`/`.yml` with OpenAPI content
11	AsciiDoc	`asciidoc` / `create file.adoc`	`asciidoc`	`.adoc`/`.asciidoc` extensions
12	PowerPoint	`pptx` / `create file.pptx`	`pptx`	`.pptx` extension
13	RSS/Atom feed	`rss` / `create feed.rss`	`rss`	`.rss`/`.atom` extensions
14	Man pages	`manpage` / `create cmd.1`	`manpage`	`.1`–`.8`/`.man` extensions
15	Confluence wiki	`confluence`	`confluence`	API or export directory
16	Notion pages	`notion`	`notion`	API or export directory
17	Slack/Discord chat	`chat`	`chat`	Export directory or API

Added

10 New Skill Source Types (17 total)

Skill Seekers now supports 17 source types — up from 7. Every new type is fully integrated into the CLI (skill-seekers <type>), create command auto-detection, unified multi-source configs, config validation, the MCP server, and the skill builder.

Jupyter Notebook — skill-seekers jupyter --notebook file.ipynb or skill-seekers create file.ipynb
- Extracts markdown cells, code cells with outputs, kernel metadata, imports, and language detection
- Handles single files and directories of notebooks; filters .ipynb_checkpoints
- Optional dependency: pip install "skill-seekers[jupyter]" (nbformat)
- Entry point: skill-seekers-jupyter
Local HTML — skill-seekers html --html-path file.html or skill-seekers create file.html
- Parses HTML using BeautifulSoup with smart main content detection (<article>, <main>, .content, largest div)
- Extracts headings, code blocks, tables (to markdown), images, links; converts inline HTML to markdown
- Handles single files and directories; supports .html, .htm, .xhtml extensions
- No extra dependencies (BeautifulSoup is a core dep)
OpenAPI/Swagger — skill-seekers openapi --spec spec.yaml or skill-seekers create spec.yaml
- Parses OpenAPI 3.0/3.1 and Swagger 2.0 specs from YAML or JSON (local files or URLs via --spec-url)
- Extracts endpoints, parameters, request/response schemas, security schemes, tags
- Resolves $ref references with circular reference protection; handles allOf/oneOf/anyOf
- Groups endpoints by tags; generates comprehensive API reference markdown
- Source detection sniffs YAML file content for openapi: or swagger: keys (avoids false positives on non-API YAML files)
- Optional dependency: pip install "skill-seekers[openapi]" (pyyaml — already a core dep, guard added for safety)
AsciiDoc — skill-seekers asciidoc --asciidoc-path file.adoc or skill-seekers create file.adoc
- Regex-based parser (no external library required) with optional asciidoc library support
- Extracts headings (= through =====), [source,lang] code blocks, |=== tables, admonitions (NOTE/TIP/WARNING/IMPORTANT/CAUTION), and include:: directives
- Converts AsciiDoc formatting to markdown; handles single files and directories
- Optional dependency: pip install "skill-seekers[asciidoc]" (asciidoc library for advanced rendering)
PowerPoint (.pptx) — skill-seekers pptx --pptx file.pptx or skill-seekers create file.pptx
- Extracts slide text, speaker notes, tables, images (with alt text), and grouped shapes
- Detects code blocks by monospace font analysis (30+ font families)
- Groups slides into sections by layout type; handles single files and directories
- Optional dependency: pip install "skill-seekers[pptx]" (python-pptx)
RSS/Atom Feeds — skill-seekers rss --feed-url <url> / --feed-path file.rss or skill-seekers create feed.rss
- Parses RSS 2.0, RSS 1.0, and Atom feeds via feedparser
- Optionally follows article links (--follow-links, default on) to scrape full page content using BeautifulSoup
- Extracts article titles, summaries, authors, dates, categories; configurable --max-articles (default 50)
- Source detection matches .rss and .atom extensions (.xml excluded to avoid false positives)
- Optional dependency: pip install "skill-seekers[rss]" (feedparser)
Man Pages — skill-seekers manpage --man-names git,curl / --man-path dir/ or skill-seekers create git.1
- Extracts man pages by running man command via subprocess or reading .1–.8/.man files directly
- Handles gzip/bzip2/xz compressed man files; strips troff/groff formatting (backspace overstriking, macros, font escapes)
- Parses structured sections (NAME, SYNOPSIS, DESCRIPTION, OPTIONS, EXAMPLES, SEE ALSO)
- Source detection uses basename heuristic to avoid false positives on log rotation files (e.g., access.log.1)
- No external dependencies (stdlib only)
Confluence — skill-seekers confluence --base-url <url> --space-key <key> or --export-path dir/
- API mode: fetches pages from Confluence REST API with pagination (atlassian-python-api)
- Export mode: parses Confluence HTML/XML export directories
- Extracts page content, code/panel/info/warning macros, page hierarchy, tables
- Optional dependency: pip install "skill-seekers[confluence]" (atlassian-python-api)
Notion — skill-seekers notion --database-id <id> / --page-id <id> or --export-path dir/
- API mode: fetches pages via Notion API with support for 20+ block types (paragraph, heading, code, callout, toggle, table, etc.)
- Export mode: parses Notion Markdown/CSV export directories
- Extracts rich text with annotations (bold, italic, code, links), 16+ property types for database entries
- Optional dependency: pip install "skill-seekers[notion]" (notion-client)
Slack/Discord Chat — skill-seekers chat --export-path dir/ or --token <token> --channel <channel>
- Slack: parses workspace JSON exports or fetches via Slack Web API (slack_sdk)
- Discord: parses DiscordChatExporter JSON or fetches via Discord HTTP API
- Extracts messages, code snippets (fenced blocks), shared URLs, threads, reactions, attachments
- Generates per-channel summaries and topic categorization
- Optional dependency: pip install "skill-seekers[chat]" (slack-sdk)

EPUB Unified Pipeline Integration

EPUB (.epub) input support via skill-seekers create book.epub or skill-seekers epub --epub book.epub
- Extracts chapters, metadata (Dublin Core), code blocks, images, and tables from EPUB 2 and EPUB 3 files
- DRM detection with clear error messages (Adobe ADEPT, Apple FairPlay, Readium LCP)
- Font obfuscation correctly identified as non-DRM
- EPUB 3 TOC bug workaround (ignore_ncx option)
- --help-epub flag for EPUB-specific help
- Optional dependency: pip install "skill-seekers[epub]" (ebooklib)
- 107 tests across 14 test classes
EPUB added to unified scraper — _scrape_epub() method, scraped_data["epub"], config validation (_validate_epub_source), and dry-run display. Previously EPUB worked standalone but was missing from multi-source configs.

Unified Skill Builder — Generic Merge System

_generic_merge() — Priority-based section merge for any combination of source types not covered by existing pairwise synthesis (docs+github, docs+pdf, etc.). Produces YAML frontmatter + source-attributed sections.
_append_extra_sources() — Appends additional source type content (e.g., Jupyter + PPTX) to pairwise-synthesized SKILL.md.
_generate_generic_references() — Generates references/<type>/index.md for any source type, with ID resolution fallback chain.
_SOURCE_LABELS dict — Human-readable labels for all 17 source types used in merge attribution.

Config Validator Expansion

17 source types in VALID_SOURCE_TYPES — All new types plus word and video now have per-type validation methods.
_validate_word_source() — Validates path field for Word documents (was previously missing).
_validate_video_source() — Validates url, path, or playlist field for video sources (was previously missing).
11 new _validate_*_source() methods — One for each new type with appropriate required-field checks.

Source Detection Improvements

7 new file extension detections in SourceDetector.detect() — .ipynb, .html/.htm, .pptx, .adoc/.asciidoc, .rss/.atom, .1–.8/.man, .yaml/.yml (with content sniffing)
_looks_like_openapi() — Content sniffing for YAML files: only classifies as OpenAPI if the file contains openapi: or swagger: key in first 20 lines (prevents false positives on docker-compose, Ansible, Kubernetes manifests, etc.)
Man page basename heuristic — .1–.8 extensions only detected as man pages if the basename has no dots (e.g., git.1 matches but access.log.1 does not)
.xml excluded from RSS detection — Too generic; only .rss and .atom trigger RSS detection

MCP Server Integration

scrape_generic tool — New MCP tool handles all 10 new source types via subprocess with per-type flag mapping
_PATH_FLAGS / _URL_FLAGS dicts — Correct flag routing for each source type (e.g., jupyter→--notebook, html→--html-path, rss→--feed-url)
GENERIC_SOURCE_TYPES tuple — Lists all 10 new types for validation
Config validation display — validate_config tool now shows source details for all new types
Tool count updated — 33 → 34 tools (scraping tools 10 → 11)

CLI Wiring

10 new CLI subcommands — jupyter, html, openapi, asciidoc, pptx, rss, manpage, confluence, notion, chat in COMMAND_MODULES
10 new argument modules — arguments/{jupyter,html,openapi,asciidoc,pptx,rss,manpage,confluence,notion,chat}.py with per-type *_ARGUMENTS dicts
10 new parser modules — parsers/{jupyter,html,openapi,asciidoc,pptx,rss,manpage,confluence,notion,chat}_parser.py with SubcommandParser implementations
create command routing — _route_generic() method for all new types with correct module names and CLI flags
10 new entry points in pyproject.toml — skill-seekers-{jupyter,html,openapi,asciidoc,pptx,rss,manpage,confluence,notion,chat}
7 new optional dependency groups in pyproject.toml — [jupyter], [asciidoc], [pptx], [confluence], [notion], [rss], [chat]
[all] group updated — Includes all 7 new optional dependencies

Sync Config Command

skill-seekers sync-config — New subcommand that crawls a docs site's navigation, diffs discovered URLs against a config's start_urls, and optionally writes the updated list back with --apply (#306)
- BFS link discovery with configurable depth (default 2), max-pages, rate-limit
- Respects url_patterns.include/exclude from config
- Supports optional nav_seed_urls config field
- Handles both unified (sources array) and legacy flat config formats
- MCP sync_config tool included
- 57 tests (39 unit + 18 E2E with local HTTP server)

Workflow & Documentation

complex-merge.yaml — New 7-stage AI-powered workflow for complex multi-source merging (source inventory → cross-reference → conflict detection → priority merge → gap analysis → synthesis → quality check)
AGENTS.md rewritten — Updated with all 17 source types, scraper pattern docs, project layout, and key pattern documentation
77 new integration tests in test_new_source_types.py — Source detection, config validation, generic merge, CLI wiring, validation, and create command routing
docs/BEST_PRACTICES.md — Comprehensive guide for creating high-quality skills: SKILL.md structure, code examples, prerequisites, troubleshooting, quality targets, and real-world Grade F to Grade A example (#206)
Documentation updated for 17 source types — 32 files updated across README, CLI reference, feature matrix, MCP reference, config format, API reference, unified scraping, multi-source guide, installation, quick-start, core concepts, user guide, FAQ, troubleshooting, architecture, and all Chinese (zh-CN) translations
README translations for 10 languages (12 total) — Added Japanese (日本語), Korean (한국어), Spanish (Español), French (Français), German (Deutsch), Portuguese (Português), Turkish (Türkçe), Arabic (العربية), Hindi (हिन्दी), and Russian (Русский) README translations with language selector bar across all versions

Performance

Pre-compiled regex and O(1) URL dedup in doc_scraper — Module-level compiled patterns, _enqueued_urls set for O(1) dedup, cached URL patterns, async error logging fix (#309)
Bisect-based line indexing in code_analyzer and dependency_analyzer — O(log n) offset_to_line() via bisect replaces O(n) count("\n") across all 10 language analyzers and all import extractors
O(n) parent class map for Python method detection — Replaces O(n²) repeated AST walks in code_analyzer
O(1) tree traversal in github_scraper — deque.popleft() replaces list pop(0)
Shared build_line_index() / offset_to_line() utilities in cli/utils.py — DRY extraction from code_analyzer and dependency_analyzer

Fixed

Config validator missing word and video dispatch — _validate_source() had no elif branches for word or video types, silently skipping validation. Added dispatch entries and _validate_word_source() / _validate_video_source() methods.
openapi_scraper.py unconditional import yaml — Would crash at import time if pyyaml not installed. Added try/except ImportError guard with YAML_AVAILABLE flag and _check_yaml_deps() helper.
asciidoc_scraper.py missing standard arguments — main() manually defined args instead of using add_asciidoc_arguments(). Refactored to use shared argument definitions + added enhancement workflow integration.
pptx_scraper.py missing standard arguments — Same issue. Refactored to use add_pptx_arguments().
chat_scraper.py missing standard arguments — Same issue. Refactored to use add_chat_arguments().
notion_scraper.py missing run_workflows call — --enhance-workflow flags were silently ignored. Added workflow runner integration.
openapi_scraper.py return type None — main() returned None instead of int. Fixed to return 0 on success, matching all other scrapers.
MCP scrape_generic_tool flag mismatch — Was passing --path/--url as generic flags, but every scraper expects its own flag name (e.g., --notebook, --html-path, --spec). All 10 source types would have failed at runtime. Fixed with per-type _PATH_FLAGS and _URL_FLAGS mappings.
Word scraper docx_id key mismatch — Unified scraper data dict used docx_id but generic reference generation looked for word_id. Added word_id alias.
main.py docstring stale — Missing all 10 new commands. Updated to list all 27 commands.
source_detector.py module docstring stale — Described only 5 source types. Updated to describe 14+ detected types.
manpage_parser.py docstring referenced wrong file — Said manpage_scraper.py but actual file is man_scraper.py. Fixed.
Parser registry test count — Updated expected count from 25 to 35 for 10 new parsers.
'Invalid IPv6 URL' error on bracket-containing URLs (#284) — URLs with square brackets (e.g., /api/[v1]/users) discovered via BFS crawl or HTML extraction bypassed the original fix in _clean_url(). Added shared sanitize_url() utility applied at every URL ingestion point. 16 new tests.
GitHub scraper 'list index out of range' on issue extraction (#269) — PyGithub's PaginatedList slicing could fail on some versions or empty repos. Replaced with itertools.islice().
Release workflow version mismatch — GitHub release showed wrong version (v3.1.3 instead of v3.2.0) because no explicit release name was set and sed regex had unescaped dots. Added explicit name/tag_name, version consistency check (tag vs pyproject.toml vs package), and empty release notes fallback.
Release workflow Python 3.10 compatibility — Version consistency check used tomllib (Python 3.11+). Replaced with grep/sed for 3.10 compatibility.
infer_categories() "tutorial" vs "tutorials" key mismatch — Guard checked 'tutorial' but wrote to 'tutorials' key, risking silent overwrites in category inference.
Flaky test_benchmark_metadata_overhead — Stabilized with 20 iterations, warm-up run, median averaging, and 200% threshold (was failing on CI with 5 iterations and mean).
CI branch protection check permanently pending — Summary job was named 'All Checks Complete' but branch protection required 'Tests'. PRs were stuck as 'Expected — Waiting for status to be reported'. Renamed job to match.

[3.2.0] - 2026-03-01

Theme: Video source support, Word document support, Pinecone adaptor, and quality improvements. 94 files changed, +23,500 lines since v3.1.3. 2,540 tests passing.

🎬 Video Tutorial Scraping Pipeline (BETA)

Complete video tutorial extraction system that converts YouTube videos and local video files into AI-consumable skills. The pipeline extracts transcripts, performs visual OCR on code editor panels, tracks code evolution across frames, and generates structured SKILL.md output.

Added

Video Pipeline Core (`skill-seekers video`)

skill-seekers video --url <youtube-url> — New CLI command for video tutorial scraping. Also supports --video-file for local files and --playlist for YouTube playlists
skill-seekers create <youtube-url> — Auto-detects YouTube URLs and routes to video scraper
video_scraper.py (~960 lines) — Main orchestrator: metadata → transcript → segmentation → visual extraction → SKILL.md generation
video_models.py (~815 lines) — 20+ dataclasses: VideoMetadata, TranscriptSegment, VideoChapter, KeyframeData, FrameSubSection, TextBlock, CodeTimeline, SetupModules, etc.
video_metadata.py (~270 lines) — YouTube metadata extraction (title, channel, views, chapters, duration) via yt-dlp; local file metadata via ffprobe
video_transcript.py (~370 lines) — Multi-source transcript extraction with 3-tier fallback: YouTube Transcript API → yt-dlp subtitles → faster-whisper local transcription
video_segmenter.py (~220 lines) — Chapter-based and time-window segmentation with configurable overlap
video_visual.py (~2,410 lines) — Visual extraction pipeline:
- Keyframe detection via scene change (scenedetect) with configurable threshold
- Frame classification (code editor, slides, terminal, browser, other)
- Panel detection — splits IDE screenshots into independent sub-sections (code, terminal, file tree)
- Per-panel OCR — Each detected panel OCR'd independently with its own bounding box
- Multi-engine OCR ensemble — EasyOCR + pytesseract for code frames (per-line confidence merge with code-token preference), EasyOCR only for non-code frames
- Parallel OCR — ThreadPoolExecutor for multi-panel frames
- Narrow panel filtering (300px min width) to skip UI chrome
- Text block tracking with spatial panel position matching across frames
- Code timeline with edit tracking (additions, modifications, deletions)
- Vision API fallback when OCR confidence < 0.5
- Tesseract circuit breaker (_tesseract_broken flag) — disables pytesseract after first failure
Audio-visual alignment — Code blocks paired with narrator transcript for context
Video-specific AI enhancement — Custom prompt for OCR denoising, code reconstruction, and tutorial narrative synthesis
Two-pass AI enhancement — Pass 1 cleans reference files (Code Timeline reconstruction from transcript context), Pass 2 generates SKILL.md from cleaned references
_ai_clean_reference() — Sends reference file to Claude to reconstruct code blocks using transcript context, fixing OCR noise before SKILL.md generation
video-tutorial.yaml workflow preset — 4-stage enhancement pipeline (OCR cleanup → language detection → tutorial synthesis → skill polish)
Video arguments — arguments/video.py with VIDEO_ARGUMENTS dict: --url, --video-file, --playlist, --vision-ocr, --keyframe-threshold, --max-keyframes, --whisper-model, --setup, etc.
Video parser — parsers/video_parser.py for unified CLI parser registry
MCP scrape_video tool — Full video scraping from MCP server with 6 visual params, setup mode, and playlist support
tests/test_video_scraper.py (197 tests) — Comprehensive coverage: models, metadata, transcript, segmenter, visual extraction, OCR, panel detection, scraper integration, CLI arguments, OCR cleaning, code filtering

Video `--setup`: GPU Auto-Detection & Dependency Installation

skill-seekers video --setup — One-command GPU auto-detection and dependency installation
- video_setup.py (~835 lines) — Complete setup orchestration module
- GPU auto-detection — Detects NVIDIA (nvidia-smi → CUDA version), AMD (rocminfo → ROCm version), or CPU-only without requiring PyTorch
- Correct PyTorch variant — Installs from the right index URL: cu124/cu121/cu118 for NVIDIA, rocm6.3/rocm6.2.4 for AMD, cpu for CPU-only
- ROCm configuration — Sets MIOPEN_FIND_MODE=FAST and HSA_OVERRIDE_GFX_VERSION for AMD GPUs
- Virtual environment detection — Warns users outside a venv with opt-in --force override
- System dependency checks — Validates tesseract and ffmpeg binaries, provides OS-specific install instructions
- Module selection — SetupModules dataclass for optional component selection (easyocr, opencv, tesseract, scenedetect, whisper)
- Base video deps always included — yt-dlp and youtube-transcript-api installed automatically
- Verification step — Post-install import checks including torch.cuda.is_available() and torch.version.hip
- Non-interactive mode — run_setup(interactive=False) for MCP server and CI/CD use
--setup early-exit — Runs before source validation (no --url required)
MCP scrape_video setup parameter — setup: bool = False in server_fastmcp.py and scraping_tools.py
create command routing — Forwards --setup to video scraper
tests/test_video_setup.py (60 tests) — GPU detection, CUDA/ROCm version mapping, installation, verification, venv checks, system deps, module selection

Microsoft Word (.docx) Support

skill-seekers word --docx <file> and skill-seekers create document.docx — Full pipeline: mammoth → HTML → BeautifulSoup → sections → SKILL.md + references/
- word_scraper.py — WordToSkillConverter class (~600 lines) with heading/code/table/image/metadata extraction
- arguments/word.py — add_word_arguments() + WORD_ARGUMENTS dict
- parsers/word_parser.py — WordParser for unified CLI parser registry
- tests/test_word_scraper.py — Comprehensive test suite (~300 lines)
.docx auto-detection in source_detector.py — Routes to word scraper
--help-word flag in create command for Word-specific help
Word support in unified scraper — _scrape_word() method for multi-source scraping
skill-seekers-word entry point in pyproject.toml
docx optional dependency group — pip install skill-seekers[docx] (mammoth + python-docx)

Other Additions

Pinecone adaptor — pinecone_adaptor.py with full upload support
video and video-full optional dependency groups in pyproject.toml
skill-seekers-video entry point in pyproject.toml
Video plan documents — 8 design documents in docs/plans/video/ (research, data models, pipeline, integration, output, testing, dependencies, overview)

Fixed

Video Pipeline OCR Quality Fixes (6)

Webcam/OTHER frames skip OCR — WEBCAM and OTHER frame types no longer get OCR'd, eliminating ~64 junk OCR results per video
_clean_ocr_line() helper — Strips leading line numbers, IDE tab bar text, Unity Inspector labels, and VS Code collapse markers from OCR output
_fix_intra_line_duplication() — Detects and removes token sequence repetition from multi-engine OCR overlap (e.g., gpublic class Card Jpublic class Card → public class Card)
_is_likely_code() filter — Reference file code fences now filtered to reject UI junk (Inspector, Hierarchy, Canvas labels) that passed frame classification
Language detection on text groups — get_text_groups() now runs LanguageDetector.detect_from_code() on each group, filling the previously-always-None detected_language field
OCR cleaning in text assembly — _assemble_structured_text() applies _clean_ocr_line() to every line before joining

Video Pipeline Fixes (15)

extract_visual_data returning 2-tuple instead of 3 — Caused ValueError crash when unpacking results
pytesseract in core deps — Moved from core dependencies to [video-full] optional group
30-min timeout for video enhancement subprocess — Previously could hang indefinitely
scrape_video_impl missing from MCP server fallback import — Added to import block
Auto-generated YouTube captions not detected — Now checks is_generated property on transcripts
--vision-ocr and --video-playlist not forwarded — create command now passes these to video scraper
Filename collision for non-ASCII video titles — Falls back to video_id when title contains non-ASCII characters
_vision_used not a proper dataclass field — Made a proper field on FrameSubSection dataclass
6 visual params missing from MCP scrape_video — Exposed keyframe_threshold, max_keyframes, whisper_model, vision_ocr, video_playlist, video_file
Missing video dep install instructions in unified scraper — Added guidance when video dependencies are not installed
MCP docstring tool counts outdated — Updated from 25→33 tools across 7 categories
Video and word commands missing from main.py docstring — Added to CLI help text
video-full exclusion from [all] deps undocumented — Added comment in pyproject.toml
Parser registry test count wrong — Updated expected count from 22→23 for video parser

Scraper & Quality Fixes

Issue #300: Selector fallback & dry-run link discovery — create https://reactflow.dev/ now finds 20+ pages (was 1):
- extract_content() extracted links after early-return → moved before
- Dry-run used main.find_all("a") instead of soup.find_all("a") → fixed
- Async dry-run had no link extraction at all → added
- get_configuration() CSS comma selector conflicted with fallback loop → removed default
- create --config with base_url config incorrectly routed to unified_scraper → now peeks at JSON
- Selector fallback duplicated in 3 places with body fallback → extracted FALLBACK_MAIN_SELECTORS constant + _find_main_content() helper
Issue #301: setup.sh fails on macOS — pip3 pointed to different Python than python3. Changed to python3 -m pip.
RAG chunking crash (AttributeError: output_dir) — converter.output_dir doesn't exist on DocToSkillConverter. Changed to Path(converter.skill_dir).
--var flag silently dropped in create routing — main.py read args.workflow_var instead of args.var
--chunk-overlap-tokens missing from package command — Wired through entire pipeline: package_skill() → adaptor.package() → format_skill_md() → _maybe_chunk_content() → RAGChunker
Chunk overlap auto-scaling — Auto-scales to max(50, chunk_tokens // 10) when chunk size is non-default
Weaviate ImportError masked by generic handler — Added except ImportError before except Exception
Hardcoded chunk defaults in 12 adaptors — Replaced 512/50 with DEFAULT_CHUNK_TOKENS/DEFAULT_CHUNK_OVERLAP_TOKENS constants
Reference file code truncation — codebase_scraper.py no longer truncates code blocks to 500 chars (5 locations)
Enhancement code block limit — summarize_reference() now uses character-budget approach instead of [:5] cap
Intro boundary code block desync — Tracks code block state to prevent splitting inside code blocks
Hardcoded python language — unified_skill_builder.py and how_to_guide_builder.py now use detected language
GitHub reference file limits removed — No more caps on issues (was 20), releases (was 10), or release bodies (was 500 chars)
GitHub scraper reference limits removed — github_scraper.py no longer caps open_issues at 20 or closed_issues at 10
PDF scraper fixes — Real API/LOCAL enhancement (was stub); removed [:3] reference file limit
Word scraper code detection — Detect mammoth monospace <p><br> blocks as code
Language detector method — Fixed detect_from_text → detect_from_code in word scraper
.docx file extension validation — Non-.docx files raise ValueError with clear message
Double _score_code_quality() call — Consolidated to single call in word scraper
--no-preserve-code renamed — Now --no-preserve-code-blocks (backward-compat alias kept)
Dead variable — Removed unused _target_lines in enhance_skill_local.py

Changed

easyocr removed from video-full optional deps — Was pulling ~2GB of NVIDIA CUDA packages regardless of GPU vendor. Now installed via --setup with correct PyTorch variant.
Video dependency error messages — video_scraper.py and video_visual.py now suggest skill-seekers video --setup as primary fix
Shared embedding methods consolidated — _generate_openai_embeddings() and _generate_st_embeddings() moved to SkillAdaptor base class, eliminating ~150 lines of duplication from chroma/weaviate/pinecone adaptors
Chunk constants centralized — DEFAULT_CHUNK_TOKENS = 512 and DEFAULT_CHUNK_OVERLAP_TOKENS = 50 in arguments/common.py, used across all 12 adaptors + rag_chunker + base + package_skill + create_command
Enhancement summarizer architecture — Character-budget approach with target_ratio for both code blocks and heading chunks

[3.1.3] - 2026-02-24

🐛 Hotfix — Explicit Chunk Flags & Argument Pipeline Cleanup

Fixed

Issue #299: skill-seekers package --target claude unrecognised argument crash — _reconstruct_argv() in main.py emits default flag values back into argv when routing subcommands. package_skill.py had a 105-line inline argparser that used different flag names to those in arguments/package.py, so forwarded flags were rejected. Fixed by replacing the inline block with a call to add_package_arguments(parser) — the single source of truth.

Changed

package_skill.py argparser refactored — Replaced ~105 lines of inline argparse duplication with a single add_package_arguments(parser) call. Flag names are now guaranteed consistent with _reconstruct_argv() output, preventing future argument-name drift.
Explicit chunk flag names — All --chunk-* flags now include unit suffixes to eliminate ambiguity between RAG tokens and streaming characters:
- --chunk-size (RAG tokens) → --chunk-tokens
- --chunk-overlap (RAG tokens) → --chunk-overlap-tokens
- --chunk (enable RAG chunking) → --chunk-for-rag
- --streaming-chunk-size (chars) → --streaming-chunk-chars
- --streaming-overlap (chars) → --streaming-overlap-chars
- --chunk-size in PDF extractor (pages) → --pdf-pages-per-chunk
setup_logging() centralized — Added setup_logging(verbose, quiet) to utils.py and removed 4 duplicate module-level logging.basicConfig() calls from doc_scraper.py, github_scraper.py, codebase_scraper.py, and unified_scraper.py

[3.1.2] - 2026-02-24

🔧 Fix `create` Command Argument Forwarding, Gemini Model, and Enhance Dispatcher

Fixed

create command argument forwarding — Universal flags (--dry-run, --verbose, --quiet, --name, --description) now work correctly across all source types. Previously, create <url> -p quick --dry-run, create owner/repo --dry-run, and create ./path --dry-run would crash because sub-scrapers didn't accept those flags
skill-seekers analyze --dry-run — Fixed _handle_analyze_command() in main.py not forwarding --dry-run, --preset, --quiet, --name, --description, --api-key, and workflow flags to codebase_scraper
Gemini model 404 errors — Replaced retired gemini-2.0-flash-exp with gemini-2.5-flash (stable GA) in the Gemini adaptor. Users attempting Gemini enhancement were getting 404 Not Found errors
skill-seekers enhance auto-detection — The documented behaviour of auto-detecting API vs LOCAL mode was never implemented. enhance now correctly routes to the platform API when a key is present: ANTHROPIC_API_KEY → Claude API, GOOGLE_API_KEY → Gemini API, OPENAI_API_KEY → OpenAI API, no key → LOCAL mode (Claude Code Max, free). Use --mode LOCAL to force local mode regardless

Added

Shared argument contract — New add_all_standard_arguments(parser) in arguments/common.py registers common + behavior + workflow args on any parser as a single call
BEHAVIOR_ARGUMENTS — Centralized --dry-run, --verbose, --quiet definitions in arguments/common.py
--dry-run for GitHub scraper — skill-seekers github --repo owner/repo --dry-run now previews the operation
--dry-run for PDF scraper — skill-seekers pdf --name test --dry-run now previews the operation
--verbose/--quiet for GitHub and PDF scrapers — Logging level control now works consistently across all scrapers
--name/--description for codebase analyzer — Custom skill name and description can now be passed to skill-seekers analyze
--mode LOCAL flag for skill-seekers enhance — Explicitly forces LOCAL mode even when API keys are present

Changed

Argument deduplication — Removed duplicated argument definitions from arguments/github.py, arguments/scrape.py, arguments/analyze.py, arguments/pdf.py; all now import shared args from arguments/common.py
create command _add_common_args() — Only forwards truly universal flags; route-specific flags (--preset, --config, --chunk-for-rag, etc.) moved to their respective route methods
codebase_scraper.py argparser — Replaced ~190 lines of inline argparser with add_analyze_arguments(parser) call

[3.1.1] - 2026-02-23

🐛 Hotfix

Fixed

create command max_pages AttributeError — Fixed crash when max_pages argument was not provided in web source routing. Uses getattr() for safe attribute access (#293, #294)

Changed

Version bump to 3.1.1

[3.1.0] - 2026-02-23

🎯 "Unified CLI & Developer Experience" — Feature Release

Theme: One command for everything. Better developer tooling. 2280+ tests passing.

Added

Unified `create` Command

Single command for all source types — auto-detects URL, GitHub repo (owner/repo), local directory, PDF file, or multi-source config JSON

skill-seekers create https://docs.react.dev/
skill-seekers create facebook/react
skill-seekers create ./my-project
skill-seekers create tutorial.pdf

Progressive help disclosure — default --help shows 13 universal flags; detailed help per source:
- --help-web, --help-github, --help-local, --help-pdf, --help-advanced, --help-all
-p shortcut for preset selection: skill-seekers create <source> -p quick|standard|comprehensive
--local-repo-path flag for specifying local clone path in create command with validation
Supports multi-source config files as input (routes to unified scraper)

Enhancement Workflow Preset System

New workflows CLI subcommand to manage enhancement workflow presets
65 bundled workflow presets shipped as YAML files in skill_seekers/workflows/:
- Core: default, minimal, security-focus, architecture-comprehensive, api-documentation
- Domain-specific: rest-api-design, graphql-schema, grpc-services, websockets-realtime, event-driven, message-queues, stream-processing
- Architecture: microservices-patterns, serverless-architecture, kubernetes-deployment, devops-deployment, terraform-guide
- Frontend: responsive-design, component-library, forms-validation, design-system, pwa-checklist, ssr-guide, deep-linking, state-management
- Quality: testing-focus, testing-frontend, performance-optimization, observability-stack, troubleshooting-guide, accessibility-a11y
- Data: database-schema, data-validation, feature-engineering, vector-databases, mlops-pipeline, model-deployment, computer-vision
- Security: encryption-guide, iam-identity, secrets-management, compliance-gdpr, auth-strategies
- Cloud: aws-services, backup-disaster-recovery
- Patterns: advanced-patterns, api-evolution, migration-guide, contribution-guide, onboarding-beginner, comparison-matrix, sdk-integration, platform-specific, cli-tooling, build-tools
- Mobile: push-notifications, offline-first, localization-i18n
- Background: background-jobs, rate-limiting, caching-strategies, webhook-guide, api-gateway
User presets stored in ~/.config/skill-seekers/workflows/
Subcommands:
- skill-seekers workflows list — List all bundled + user workflows with descriptions
- skill-seekers workflows show <name> — Print YAML content of a workflow
- skill-seekers workflows copy <name> [name ...] — Copy bundled workflow(s) to user dir
- skill-seekers workflows add <file.yaml> [file ...] — Install custom YAML file(s) into user dir
- skill-seekers workflows remove <name> [name ...] — Delete user workflow(s)
- skill-seekers workflows validate <name|path> — Parse and validate a workflow
copy, add, remove all accept multiple names/files in one command (partial-failure: continues processing, returns non-zero if any item fails)
New entry point: skill-seekers-workflows

Multiple `--enhance-workflow` Flags from CLI

Chain workflows in a single command: skill-seekers create <source> --enhance-workflow security-focus --enhance-workflow minimal
Supported across all scrapers: scrape, github, analyze, pdf, unified

Smart Enhancement Dispatcher (`skill-seekers enhance`)

Auto-routes to API mode (Claude/Gemini/OpenAI) when API key is available, LOCAL mode (Claude Code CLI) otherwise
Decision priority: --target flag → config default_agent → env vars (ANTHROPIC_API_KEY → claude, GOOGLE_API_KEY → gemini, OPENAI_API_KEY → openai) → LOCAL fallback
Blocks LOCAL mode when running as root (Docker/VPS) with clear error message + API mode instructions (fixes #286, #289)
New flags: --target, --api-key, --dry-run, --interactive-enhancement

Unified Document Parser System

New parsers/extractors.py module with RstParser, MarkdownParser classes
ReStructuredText (RST) support — parses class references, code blocks, tables, cross-references
Shared parse_document() factory function for RST/Markdown/PDF input
Integrated into documentation extraction pipeline for richer content
ContentBlockType and CrossRefType enums for structured parsing output

Local Source Support in Unified Scraper

"type": "local" source type in unified config JSONs — analyze local codebases alongside web/GitHub/PDF sources
--local-repo-path CLI flag in unified scraper for per-source path override

CLI Flag Parity Across All Commands

analyze, pdf, and unified commands now have full flag parity with scrape/github:
- --api-key on pdf and unified
- --enhance-level on unified
- --dry-run on analyze
- All workflow flags (--enhance-workflow, --enhance-stage, --var, --workflow-dry-run) on analyze
Workflow JSON config fields (workflows, workflow_stages, workflow_vars) now merged with CLI flags in unified scraper

Fixed

Percent-encode brackets in llms.txt URLs — prevent "Invalid IPv6 URL" errors when scraping sites with bracket characters (fixes #284)
Platform-appropriate config paths on Windows — use %APPDATA% instead of ~/.config (fixes #283)
create command multi-source config — now correctly routes to unified scraper when input is a .json config file
create command _add_common_args() — correctly forwards each --enhance-workflow value as a separate flag to sub-scrapers (previously collapsed list to single string, causing workflows to be ignored)
_extract_markdown_content — filter out bare h1 headings and short stub paragraphs that polluted extracted content
Godot unified config language names — corrected gdscript/gds to proper names in godot_unified.json
Python 3.10 type union compatibility — use Optional[X] instead of X | None in forward-reference positions
_route_config in unified scraper — correctly handles all source types when routing config-driven scraping
CONFIG_ARGUMENTS — added to ensure unified CLI has full argument visibility for config-based sources
Test suite isolation — test_swift_detection.py now saves/restores sys.modules and parent package attributes; prevents @patch decorators in downstream files from targeting stale module objects
Python 3.14 chromadb compatibility — catch pydantic.v1.errors.ConfigError (not just ImportError) when chromadb is installed
langchain import path — updated langchain.schema → langchain_core.documents for langchain 1.x
Removed legacy sys.path.insert() calls from codebase_scraper.py, doc_scraper.py, enhance_skill.py, enhance_skill_local.py, estimate_pages.py, install_skill.py (unnecessary with pip install -e .)
Benchmark timing threshold — relaxed metadata overhead assertion from 10% to 50% for CI runner variability

Changed

Enhancement flags consolidated — --enhance-level (0-3) replaces three separate flags (--enhance, --enhance-local, --api-key). Old flags still accepted with deprecation warnings until v4.0.0
workflows copy/add/remove now accept multiple names/files in one invocation
pyproject.toml — PyYAML added as core dependency (required by workflow preset management); langchain and llama-index added as dependencies; MCP version requirement updated to >=1.25

Tests

2280+ tests passing (2158 non-MCP + ~122 MCP, up from 1852 in v3.0.0), 11 skipped (external services), 0 failures
Added TestAnalyzeWorkflowFlags, TestUnifiedCLIArguments, TestPDFCLIArguments classes
Added tests/test_mcp_workflow_tools.py — 5 MCP workflow tool tests
Added tests/test_unified_scraper_orchestration.py — UnifiedScraper orchestration tests
Removed @unittest.skip from gemini/openai/claude adaptor tests that were ready
Removed @requires_github from 5 unified_analyzer tests that fully mock their dependencies
Macros-specific tests now use @patch(sys.platform) instead of runtime skipTest() for platform portability

Config Repository (skill-seekers-configs)

178 production configs reviewed and enhanced across all 22 categories — brought to v1.1.0 quality standard
Removed all max_pages fields from production configs (deprecated, defaults apply automatically)
Fixed outdated URLs: astro.json (Astro v3 restructure: /en/core-concepts/ → /en/basics/), laravel.json (11.x → 12.x throughout)
Fixed structural bug in httpx_comprehensive.json — url_patterns, categories, rate_limit moved from top-level into sources[0] (required for unified format)
Removed hash-fragment start_urls from zod.json (scrapers don't follow ?id= anchors)
Improved category/selector quality across all 22 categories: 5-13 categories per config, 3-6 keywords each, semantic selector fallback chains
README.md: corrected config count from outdated "50+" to accurate 178 production / 182 total; all category counts verified
CONTRIBUTING.md, QUALITY_GUIDELINES.md, AGENTS.md: aligned with production standards; removed all max_pages guidance
scripts/validate-config.py: fixed two bugs — unified config categories lookup (was always reporting "no categories" for multi-source configs) and max_pages warning logic (was warning when absent, now correctly warns when present)
Deleted .github/ISSUE_TEMPLATE/submit-config.md (old duplicate of submit-config.yml with outdated content)

[3.0.0] - 2026-02-10

🚀 "Universal Intelligence Platform" - Major Release

Theme: Transform any documentation into structured knowledge for any AI system.

This is our biggest release ever! v3.0.0 establishes Skill Seekers as the universal documentation preprocessor for the entire AI ecosystem - from RAG pipelines to AI coding assistants to Claude skills.

Highlights

🚀 16 platform adaptors (up from 4 in v2.x)
🛠️ 26 MCP tools (up from 9)
✅ 1,852 tests passing (up from 700+)
☁️ Cloud storage support (S3, GCS, Azure)
🔄 CI/CD ready (GitHub Action + Docker)
📦 12 example projects for every integration
📚 18 integration guides complete

Added - Platform Adaptors (16 Total)

RAG & Vector Databases (8)

LangChain (--format langchain) - Output LangChain Document objects
LlamaIndex (--format llama-index) - Output LlamaIndex TextNode objects
Chroma (--format chroma) - Direct ChromaDB integration
FAISS (--format faiss) - Facebook AI Similarity Search
Haystack (--format haystack) - Deepset Haystack pipelines
Qdrant (--format qdrant) - Qdrant vector database
Weaviate (--format weaviate) - Weaviate vector search
Pinecone-ready (--target markdown) - Markdown format ready for Pinecone

AI Platforms (3)

Claude (--target claude) - Claude AI skills (ZIP + YAML)
Gemini (--target gemini) - Google Gemini skills (tar.gz)
OpenAI (--target openai) - OpenAI ChatGPT (ZIP + Vector Store)

AI Coding Assistants (4)

Cursor (--target claude + .cursorrules) - Cursor IDE integration
Windsurf (--target claude + .windsurfrules) - Windsurf/Codeium
Cline (--target claude + .clinerules) - VS Code extension
Continue.dev (--target claude) - Universal IDE support

Generic (1)

Markdown (--target markdown) - Generic ZIP export

Added - MCP Tools (26 Total)

Config Tools (3)

generate_config - Generate scraping configuration
list_configs - List available preset configs
validate_config - Validate config JSON structure

Scraping Tools (8)

estimate_pages - Estimate page count before scraping
scrape_docs - Scrape documentation websites
scrape_github - Scrape GitHub repositories
scrape_pdf - Extract from PDF files
scrape_codebase - Analyze local codebases
detect_patterns - Detect design patterns in code
extract_test_examples - Extract usage examples from tests
build_how_to_guides - Build how-to guides from code

Packaging Tools (4)

package_skill - Package skill for target platform
upload_skill - Upload to LLM platform
enhance_skill - AI-powered enhancement
install_skill - One-command complete workflow

Source Tools (5)

fetch_config - Fetch config from remote source
submit_config - Submit config for approval
add_config_source - Add Git config source
list_config_sources - List config sources
remove_config_source - Remove config source

Splitting Tools (2)

split_config - Split large configs
generate_router - Generate router skills

Vector DB Tools (4)

export_to_weaviate - Export to Weaviate
export_to_chroma - Export to ChromaDB
export_to_faiss - Export to FAISS
export_to_qdrant - Export to Qdrant

Added - Cloud Storage

Upload skills directly to cloud storage:

AWS S3 - skill-seekers cloud upload --provider s3 --bucket my-bucket
Google Cloud Storage - skill-seekers cloud upload --provider gcs --bucket my-bucket
Azure Blob Storage - skill-seekers cloud upload --provider azure --container my-container

Features:

Upload/download directories
List files with metadata
Check file existence
Generate presigned URLs
Cloud-agnostic interface

Added - CI/CD Support

GitHub Action

- uses: skill-seekers/action@v1
  with:
    config: configs/react.json
    format: langchain

Features:

Auto-update on doc changes
Matrix builds for multiple frameworks
Scheduled updates
Caching for faster runs

Docker

docker run -v $(pwd):/data skill-seekers:latest scrape --config /data/config.json

Added - Production Infrastructure

Helm Charts - Kubernetes deployment
Docker Compose - Local vector DB stack
Monitoring - Sentry integration, sync monitoring
Benchmarking - Performance testing framework

Added - 12 Example Projects

Complete working examples for every integration:

langchain-rag-pipeline - React docs → LangChain → Chroma
llama-index-query-engine - Vue docs → LlamaIndex
pinecone-upsert - Documentation → Pinecone
chroma-example - Full ChromaDB workflow
faiss-example - FAISS index building
haystack-pipeline - Haystack RAG pipeline
qdrant-example - Qdrant vector DB
weaviate-example - Weaviate integration
cursor-react-skill - React skill for Cursor
windsurf-fastapi-context - FastAPI for Windsurf
cline-django-assistant - Django assistant for Cline
continue-dev-universal - Universal IDE context

Quality Metrics

✅ 1,852 tests across 100 test files
✅ 58,512 lines of Python code
✅ 80+ documentation files
✅ 100% test coverage for critical paths
✅ CI/CD on every commit

Fixed

URL Conversion Bug with Anchor Fragments (Issue #277)

Critical Bug Fix: Fixed 404 errors when scraping documentation with anchor links
- Problem: URLs with anchor fragments (e.g., #synchronous-initialization) were malformed
  - Incorrect: https://example.com/docs/api#method/index.html.md ❌
  - Correct: https://example.com/docs/api/index.html.md ✅
- Root Cause: _convert_to_md_urls() didn't strip anchor fragments before appending /index.html.md
- Solution: Parse URLs with urllib.parse to remove fragments and deduplicate base URLs
- Impact: Prevents duplicate requests for the same page with different anchors
- Additional Fix: Changed .md detection from ".md" in url to url.endswith('.md')
  - Prevents false matches on URLs like /cmd-line or /AMD-processors
Test Coverage: 12 comprehensive tests covering all edge cases
- Anchor fragment stripping
- Deduplication of multiple anchors on same URL
- Query parameter preservation
- Trailing slash handling
- Real-world MikroORM case validation
- 54/54 tests passing (42 existing + 12 new)
Reported by: @devjones via Issue #277

Added

Extended Language Detection (NEW)

7 New Programming Languages: Dart, Scala, SCSS, SASS, Elixir, Lua, Perl
- Pattern-based detection with confidence scoring (0.6-0.8+ thresholds)
- 70 regex patterns prioritizing unique identifiers (weight 5)
- Framework-specific patterns:
  - Dart: Flutter widgets (StatelessWidget, StatefulWidget, Widget build())
  - Scala: Pattern matching (case class, trait, match {})
  - SCSS: Preprocessor features ($variables, @mixin, @include, @extend)
  - SASS: Indented syntax (=mixin, +include, $variables)
  - Elixir: Functional patterns (defmodule, def ... do, pipe operator |>)
  - Lua: Game scripting (local, repeat...until, ~=, elseif)
  - Perl: Text processing (my $, use strict, sub, chomp, regex =~)
- Comprehensive test coverage: 7 new tests, 30/30 passing (100%)
- False positive prevention: Unique identifiers (weight 5) + confidence thresholds
- No regressions: All existing language detection tests still pass
- Total language support: Now 27+ programming languages
- Credit: Contributed by @PaawanBarach via PR #275

Multi-Agent Support for Local Enhancement (NEW)

Multiple Coding Agent Support: Choose your preferred local coding agent for SKILL.md enhancement
- Claude Code (default): Claude Code CLI with --dangerously-skip-permissions
- Codex CLI: OpenAI Codex CLI with --full-auto and --skip-git-repo-check
- Copilot CLI: GitHub Copilot CLI (gh copilot chat)
- OpenCode CLI: OpenCode CLI
- Custom agents: Use any CLI tool with --agent custom --agent-cmd "command {prompt_file}"
CLI Arguments: New flags for agent selection
- --agent: Choose agent (claude, codex, copilot, opencode, custom)
- --agent-cmd: Override command template for custom agents
Environment Variables: CI/CD friendly configuration
- SKILL_SEEKER_AGENT: Default agent to use
- SKILL_SEEKER_AGENT_CMD: Default command template for custom agents
Security First: Custom command validation
- Blocks dangerous shell characters (;, &, |, $, `, \n, \r)
- Validates executable exists in PATH
- Safe parsing with shlex.split()
Dual Input Modes: Supports both file-based and stdin-based agents
- File-based: Uses {prompt_file} placeholder (Claude, custom agents)
- Stdin-based: Pipes prompt via stdin (Codex CLI)
Backward Compatible: Claude Code remains the default, no breaking changes
Comprehensive Tests: 13 new tests covering all agent types and security validation
Agent Normalization: Smart alias handling (e.g., "claude-code" → "claude")
Credit: Contributed by @rovo79 (Robert Dean) via PR #270

C3.10: Signal Flow Analysis for Godot Projects (NEW)

Complete Signal Flow Analysis System: Analyze event-driven architectures in Godot game projects
- Signal declaration extraction (signal keyword detection)
- Connection mapping (.connect() calls with targets and methods)
- Emission tracking (.emit() and emit_signal() calls)
- 208 signals, 634 connections, and 298 emissions detected in test project (Cosmic Idler)
- Signal density metrics (signals per file)
- Event chain detection (signals triggering other signals)
- Output: signal_flow.json, signal_flow.mmd (Mermaid diagram), signal_reference.md
Signal Pattern Detection: Three major patterns identified
- EventBus Pattern (0.90 confidence): Centralized signal hub in autoload
- Observer Pattern (0.85 confidence): Multi-observer signals (3+ listeners)
- Event Chains (0.80 confidence): Cascading signal propagation
Signal-Based How-To Guides (C3.10.1): AI-generated usage guides
- Step-by-step guides (Connect → Emit → Handle)
- Real code examples from project
- Common usage locations
- Parameter documentation
- Output: signal_how_to_guides.md (10 guides for Cosmic Idler)

Godot Game Engine Support

Comprehensive Godot File Type Support: Full analysis of Godot 4.x projects
- GDScript (.gd): 265 files analyzed in test project
- Scene files (.tscn): 118 scene files
- Resource files (.tres): 38 resource files
- Shader files (.gdshader, .gdshaderinc): 9 shader files
- C# integration: Phantom Camera addon (13 files)
GDScript Language Support: Complete GDScript parsing with regex-based extraction
- Dependency extraction: preload(), load(), extends patterns
- Test framework detection: GUT, gdUnit4, WAT
- Test file patterns: test_*.gd, *_test.gd
- Signal syntax: signal, .connect(), .emit()
- Export decorators: @export, @onready
- Test decorators: @test (gdUnit4)
Game Engine Framework Detection: Improved detection for Unity, Unreal, Godot
- Godot markers: project.godot, .godot directory, .tscn, .tres, .gd files
- Unity markers: Assembly-CSharp.csproj, UnityEngine.dll, ProjectSettings/ProjectVersion.txt
- Unreal markers: .uproject, Source/, Config/DefaultEngine.ini
- Fixed false positive Unity detection (was using generic "Assets" keyword)
GDScript Test Extraction: Extract usage examples from Godot test files
- 396 test cases extracted from 20 GUT test files in test project
- Patterns: instantiation (preload().new(), load().new()), assertions (assert_eq, assert_true), signals
- GUT framework: extends GutTest, func test_*(), add_child_autofree()
- Test categories: instantiation, assertions, signal connections, setup/teardown
- Real code examples from production test files

C3.9: Project Documentation Extraction

Markdown Documentation Extraction: Automatically extracts and categorizes all .md files from projects
- Smart categorization by folder/filename (overview, architecture, guides, workflows, features, etc.)
- Processing depth control: surface (raw copy), deep (parse+summarize), full (AI-enhanced)
- AI enhancement (level 2+) adds topic extraction and cross-references
- New "📖 Project Documentation" section in SKILL.md
- Output to references/documentation/ organized by category
- Default ON, use --skip-docs to disable
- 15 new tests for documentation extraction features

Granular AI Enhancement Control

--enhance-level Flag: Fine-grained control over AI enhancement (0-3)
- Level 0: No AI enhancement (default)
- Level 1: SKILL.md enhancement only (fast, high value)
- Level 2: SKILL.md + Architecture + Config + Documentation
- Level 3: Full enhancement (patterns, tests, config, architecture, docs)
Config Integration: default_enhance_level setting in ~/.config/skill-seekers/config.json
MCP Support: All MCP tools updated with enhance_level parameter
Independent from --comprehensive: Enhancement level is separate from feature depth

C# Language Support

C# Test Example Extraction: Full support for C# test frameworks
- Language alias mapping (C# → csharp, C++ → cpp)
- NUnit, xUnit, MSTest test framework patterns
- Mock pattern support (NSubstitute, Moq)
- Zenject dependency injection patterns
- Setup/teardown method extraction
- 2 new tests for C# extraction features

Performance Optimizations

Parallel LOCAL Mode AI Enhancement: 6-12x faster with ThreadPoolExecutor
- Concurrent workers: 3 (configurable via local_parallel_workers)
- Batch processing: 20 patterns per Claude CLI call (configurable via local_batch_size)
- Significant speedup for large codebases
Config Settings: New ai_enhancement section in config
- local_batch_size: Patterns per CLI call (default: 20)
- local_parallel_workers: Concurrent workers (default: 3)

UX Improvements

Auto-Enhancement: SKILL.md automatically enhanced when using --enhance or --comprehensive
- No need for separate skill-seekers enhance command
- Seamless one-command workflow
- 10-minute timeout for large codebases
- Graceful fallback with retry instructions on failure
LOCAL Mode Fallback: All AI enhancements now fall back to LOCAL mode when no API key is set
- Applies to: pattern enhancement (C3.1), test examples (C3.2), architecture (C3.7)
- Uses Claude Code CLI instead of failing silently
- Better UX: "Using LOCAL mode (Claude Code CLI)" instead of "AI disabled"
Support for custom Claude-compatible API endpoints via ANTHROPIC_BASE_URL environment variable
Compatibility with GLM-4.7 and other Claude-compatible APIs across all AI enhancement features

Changed

All AI enhancement modules now respect ANTHROPIC_BASE_URL for custom endpoints
Updated documentation with GLM-4.7 configuration examples
Rewritten LOCAL mode in config_enhancer.py to use Claude CLI properly with explicit output file paths
Updated MCP scrape_codebase_tool with skip_docs and enhance_level parameters
Updated CLAUDE.md with C3.9 documentation extraction feature
Increased default batch size from 5 to 20 patterns for LOCAL mode

Fixed

C# Test Extraction: Fixed "Language C# not supported" error with language alias mapping
Config Type Field Mismatch: Fixed KeyError in config_enhancer.py by supporting both "type" and "config_type" fields
LocalSkillEnhancer Import: Fixed incorrect import and method call in main.py (SkillEnhancer → LocalSkillEnhancer)
Code Quality: Fixed 4 critical linter errors (unused imports, variables, arguments, import sorting)

Godot Game Engine Fixes

GDScript Dependency Extraction: Fixed 265+ "Syntax error in *.gd" warnings (commit 3e6c448)
- GDScript files were incorrectly routed to Python AST parser
- Created dedicated _extract_gdscript_imports() with regex patterns
- Now correctly parses preload(), load(), extends patterns
- Result: 377 dependencies extracted with 0 warnings
Framework Detection False Positive: Fixed Unity detection on Godot projects (commit 50b28fe)
- Was detecting "Unity" due to generic "Assets" keyword in comments
- Changed Unity markers to specific files: Assembly-CSharp.csproj, UnityEngine.dll, Library/
- Now correctly detects Godot via project.godot, .godot directory
Circular Dependencies: Fixed self-referential cycles (commit 50b28fe)
- 3 self-loop warnings (files depending on themselves)
- Added target != file_path check in dependency graph builder
- Result: 0 circular dependencies detected
GDScript Test Discovery: Fixed 0 test files found in Godot projects (commit 50b28fe)
- Added GDScript test patterns: test_*.gd, *_test.gd
- Added GDScript to LANGUAGE_MAP
- Result: 32 test files discovered (20 GUT files with 396 tests)
GDScript Test Extraction: Fixed "Language GDScript not supported" warning (commit c826690)
- Added GDScript regex patterns to PATTERNS dictionary
- Patterns: instantiation (preload().new()), assertions (assert_eq), signals (.connect())
- Result: 22 test examples extracted successfully
Config Extractor Array Handling: Fixed JSON/YAML array parsing (commit fca0951)
- Error: 'list' object has no attribute 'items' on root-level arrays
- Added isinstance checks for dict/list/primitive at root
- Result: No JSON array errors, save.json parsed correctly
Progress Indicators: Fixed missing progress for small batches (commit eec37f5)
- Progress only shown every 5 batches, invisible for small jobs
- Modified condition to always show for batches < 10
- Result: "Progress: 1/2 batches completed" now visible

Other Fixes

C# Test Extraction: Fixed "Language C# not supported" error with language alias mapping
Config Type Field Mismatch: Fixed KeyError in config_enhancer.py by supporting both "type" and "config_type" fields
LocalSkillEnhancer Import: Fixed incorrect import and method call in main.py (SkillEnhancer → LocalSkillEnhancer)
Code Quality: Fixed 4 critical linter errors (unused imports, variables, arguments, import sorting)

Tests

GDScript Test Extraction Test: Added comprehensive test case for GDScript GUT/gdUnit4 framework
- Tests player instantiation with preload() and load()
- Tests signal connections and emissions
- Tests gdUnit4 @test annotation syntax
- Tests game state management patterns
- 4 test functions with 60+ lines of GDScript code
- Validates extraction of instantiations, assertions, and signal patterns

Removed

Removed client-specific documentation files from repository

[2.7.4] - 2026-01-22

🔧 Bug Fix - Language Selector Links

This patch release fixes the broken Chinese language selector link that appeared on PyPI and other non-GitHub platforms.

Fixed

Broken Language Selector Links on PyPI
- Issue: Chinese language link used relative URL (README.zh-CN.md) which only worked on GitHub
- Impact: Users on PyPI clicking "简体中文" got 404 errors
- Solution: Changed to absolute GitHub URL (https://github.com/yusufkaraaslan/Skill_Seekers/blob/main/README.zh-CN.md)
- Result: Language selector now works on PyPI, GitHub, and all platforms
- Files Fixed: README.md, README.zh-CN.md

Technical Details

Why This Happened:

PyPI displays README.md but doesn't include README.zh-CN.md in the package
Relative links break when README is rendered outside GitHub repository context
Absolute GitHub URLs work universally across all platforms

Impact:

✅ Chinese language link now accessible from PyPI
✅ Consistent experience across all platforms
✅ Better user experience for Chinese developers

[2.7.3] - 2026-01-21

🌏 International i18n Release

This documentation release adds comprehensive Chinese language support, making Skill Seekers accessible to the world's largest developer community.

Added

🇨🇳 Chinese (Simplified) README Translation (#260)
- Complete 1,962-line translation of all documentation (README.zh-CN.md)
- Language selector badges in both English and Chinese READMEs
- Machine translation disclaimer with invitation for community improvements
- GitHub issue #260 created for community review and contributions
- Impact: Makes Skill Seekers accessible to 1+ billion Chinese speakers
📦 PyPI Metadata Internationalization
- Updated package description to highlight Chinese documentation availability
- Added i18n-related keywords: "i18n", "chinese", "international"
- Added Natural Language classifiers: English and Chinese (Simplified)
- Added direct link to Chinese README in project URLs
- Impact: Better discoverability on PyPI for Chinese developers

Why This Matters

Market Reach: Addresses existing Chinese traffic and taps into world's largest developer community
Discoverability: Better indexing on Chinese search engines (Baidu, Gitee, etc.)
User Experience: Native language documentation lowers barrier to entry
Community Growth: Opens contribution opportunities from Chinese developers
Competitive Edge: Most similar tools don't offer Chinese documentation

Community Engagement

Chinese developers are invited to improve the translation quality:

Review issue: https://github.com/yusufkaraaslan/Skill_Seekers/issues/260
Translation guidelines provided for technical accuracy and natural expression
All contributions welcome and appreciated

[2.7.2] - 2026-01-21

🚨 Critical CLI Bug Fixes

This hotfix release resolves 4 critical CLI bugs reported in issues #258 and #259 that prevented core commands from working correctly.

Fixed

Issue #258: install --config command fails with unified scraper (#258)
- Root Cause: unified_scraper.py missing --fresh and --dry-run argument definitions
- Solution: Added both flags to unified_scraper argument parser and main.py dispatcher
- Impact: skill-seekers install --config react now works without "unrecognized arguments" error
- Files Fixed: src/skill_seekers/cli/unified_scraper.py, src/skill_seekers/cli/main.py
Issue #259 (Original): scrape command doesn't accept URL and --max-pages (#259)
- Root Cause: No positional URL argument or --max-pages flag support
- Solution: Added positional URL argument and --max-pages flag with safety warnings
- Impact: skill-seekers scrape https://example.com --max-pages 50 now works
- Safety Warnings:
  - ⚠️ Warning if max-pages > 1000 (may take hours)
  - ⚠️ Warning if max-pages < 10 (incomplete skill)
- Files Fixed: src/skill_seekers/cli/doc_scraper.py, src/skill_seekers/cli/main.py
Issue #259 (Comment A): Version shows 2.7.0 instead of actual version (#259)
- Root Cause: Hardcoded version string in main.py
- Solution: Import __version__ from __init__.py dynamically
- Impact: skill-seekers --version now shows correct version (2.7.2)
- Files Fixed: src/skill_seekers/cli/main.py
Issue #259 (Comment B): PDF command shows empty "Error: " message (#259)
- Root Cause: Exception handler didn't handle empty exception messages
- Solution:
  - Improved exception handler to show exception type if message is empty
  - Added proper error handling with context-specific messages
  - Added traceback support in verbose mode
- Impact: PDF errors now show clear messages like "Error: RuntimeError occurred" instead of just "Error: "
- Files Fixed: src/skill_seekers/cli/main.py, src/skill_seekers/cli/pdf_scraper.py

Testing

✅ Verified skill-seekers install --config react --dry-run works
✅ Verified skill-seekers scrape https://tailwindcss.com/docs/installation --max-pages 50 works
✅ Verified skill-seekers --version shows "2.7.2"
✅ Verified PDF errors show proper messages
✅ All 202 tests passing

[2.7.1] - 2026-01-18

🚨 Critical Bug Fix - Config Download 404 Errors

This hotfix release resolves a critical bug causing 404 errors when downloading configs from the API.

Fixed

Critical: Config download 404 errors - Fixed bug where code was constructing download URLs manually instead of using the download_url field from the API response
- Root Cause: Code was building f"{API_BASE_URL}/api/download/{config_name}.json" which failed when actual URLs differed (CDN URLs, version-specific paths)
- Solution: Changed to use config_info.get("download_url") from API response in both MCP server implementations
- Files Fixed:
  - src/skill_seekers/mcp/tools/source_tools.py (FastMCP server)
  - src/skill_seekers/mcp/server_legacy.py (Legacy server)
- Impact: Fixes all config downloads from skillseekersweb.com API and private Git repositories
- Reported By: User testing skill-seekers install --config godot --unlimited
- Testing: All 15 source tools tests pass, all 8 fetch_config tests pass

[2.7.0] - 2026-01-18

🔐 Smart Rate Limit Management & Multi-Token Configuration

This minor feature release introduces intelligent GitHub rate limit handling, multi-profile token management, and comprehensive configuration system. Say goodbye to indefinite waits and confusing token setup!

Added

🎯 Multi-Token Configuration System - Flexible GitHub token management with profiles
- Secure config storage at ~/.config/skill-seekers/config.json with 600 permissions
- Multiple GitHub profiles support (personal, work, OSS, etc.)
  - Per-profile rate limit strategies: prompt, wait, switch, fail
  - Configurable timeout per profile (default: 30 minutes)
  - Auto-detection and smart fallback chain
  - Profile switching when rate limited
- API key management for Claude, Gemini, OpenAI
  - Environment variable fallback (ANTHROPIC_API_KEY, GOOGLE_API_KEY, OPENAI_API_KEY)
  - Config file storage with secure permissions
- Progress tracking for resumable jobs
  - Auto-save at configurable intervals (default: 60 seconds)
  - Job metadata: command, progress, checkpoints, timestamps
  - Stored at ~/.local/share/skill-seekers/progress/
- Auto-cleanup of old progress files (default: 7 days, configurable)
- First-run experience with welcome message and quick setup
- ConfigManager class with singleton pattern for global access
🧙 Interactive Configuration Wizard - Beautiful terminal UI for easy setup
- Main menu with 7 options:
  1. GitHub Token Setup
  2. API Keys (Claude, Gemini, OpenAI)
  3. Rate Limit Settings
  4. Resume Settings
  5. View Current Configuration
  6. Test Connections
  7. Clean Up Old Progress Files
- GitHub token management:
  - Add/remove profiles with descriptions
  - Set default profile
  - Browser integration - opens GitHub token creation page
  - Token validation with format checking (ghp_, github_pat_)
  - Strategy selection per profile
- API keys setup with browser integration for each provider
- Connection testing to verify tokens and API keys
- Configuration display with current status and sources
- CLI commands:
  - skill-seekers config - Main menu
  - skill-seekers config --github - Direct to GitHub setup
  - skill-seekers config --api-keys - Direct to API keys
  - skill-seekers config --show - Show current config
  - skill-seekers config --test - Test connections
🚦 Smart Rate Limit Handler - Intelligent GitHub API rate limit management
- Upfront warning about token status (60/hour vs 5000/hour)
- Real-time detection of rate limits from GitHub API responses
  - Parses X-RateLimit-* headers
  - Detects 403 rate limit errors
  - Calculates reset time from timestamps
- Live countdown timers with progress display
- Automatic profile switching - tries next available profile when rate limited
- Four rate limit strategies:
  - prompt - Ask user what to do (default, interactive)
  - wait - Auto-wait with countdown timer
  - switch - Automatically try another profile
  - fail - Fail immediately with clear error
- Non-interactive mode for CI/CD (fail fast, no prompts)
- Configurable timeouts per profile (prevents indefinite waits)
- RateLimitHandler class with strategy pattern
- Integration points: GitHub fetcher, GitHub scraper
📦 Resume Command - Resume interrupted scraping jobs
- List resumable jobs with progress details:
  - Job ID, started time, command
  - Current phase and file counts
  - Last updated timestamp
- Resume from checkpoints (skeleton implemented, ready for integration)
- Auto-cleanup of old jobs (respects config settings)
- CLI commands:
  - skill-seekers resume --list - List all resumable jobs
  - skill-seekers resume <job-id> - Resume specific job
  - skill-seekers resume --clean - Clean up old jobs
- Progress storage at ~/.local/share/skill-seekers/progress/<job-id>.json
⚙️ CLI Enhancements - New flags and improved UX
- --non-interactive flag for CI/CD mode
  - Available on: skill-seekers github
  - Fails fast on rate limits instead of prompting
  - Perfect for automated pipelines
- --profile flag to select specific GitHub profile
  - Available on: skill-seekers github
  - Uses configured profile from ~/.config/skill-seekers/config.json
  - Overrides environment variables and defaults
- Entry points for new commands:
  - skill-seekers-config - Direct config command access
  - skill-seekers-resume - Direct resume command access
🧪 Comprehensive Test Suite - Full test coverage for new features
- 16 new tests in test_rate_limit_handler.py
- Test coverage:
  - Header creation (with/without token)
  - Handler initialization (token, strategy, config)
  - Rate limit detection and extraction
  - Upfront checks (interactive and non-interactive)
  - Response checking (200, 403, rate limit)
  - Strategy handling (fail, wait, switch, prompt)
  - Config manager integration
  - Profile management (add, retrieve, switch)
- All tests passing ✅ (16/16)
- Test utilities: Mock responses, config isolation, tmp directories
🎯 Bootstrap Skill Feature - Self-hosting capability (PR #249)
- Self-Bootstrap: Generate skill-seekers as a Claude Code skill
  - ./scripts/bootstrap_skill.sh - One-command bootstrap
  - Combines manual header with auto-generated codebase analysis
  - Output: output/skill-seekers/ ready for Claude Code
  - Install: cp -r output/skill-seekers ~/.claude/skills/
- Robust Frontmatter Detection:
  - Dynamic YAML frontmatter boundary detection (not hardcoded line counts)
  - Fallback to line 6 if frontmatter not found
  - Future-proof against frontmatter field additions
- SKILL.md Validation:
  - File existence and non-empty checks
  - Frontmatter delimiter presence
  - Required fields validation (name, description)
  - Exit with clear error messages on validation failures
- Comprehensive Error Handling:
  - UV dependency check with install instructions
  - Permission checks for output directory
  - Graceful degradation on missing header file
🔧 MCP Now Optional - User choice for installation profile
- CLI Only: pip install skill-seekers - No MCP dependencies
- MCP Integration: pip install skill-seekers[mcp] - Full MCP support
- All Features: pip install skill-seekers[all] - Everything enabled
- Lazy Loading: Graceful failure with helpful error messages when MCP not installed
- Interactive Setup Wizard:
  - Shows all installation options on first run
  - Stored at ~/.config/skill-seekers/.setup_shown
  - Accessible via skill-seekers-setup command
- Entry Point: skill-seekers-setup for manual access
🧪 E2E Testing for Bootstrap - Comprehensive end-to-end tests
- 6 core tests verifying bootstrap workflow:
  - Output structure creation
  - Header prepending
  - YAML frontmatter validation
  - Line count sanity checks
  - Virtual environment installability
  - Platform adaptor compatibility
- Pytest markers: @pytest.mark.e2e, @pytest.mark.venv, @pytest.mark.slow
- Execution modes:
  - Fast tests: pytest -k "not venv" (~2-3 min)
  - Full suite: pytest -m "e2e" (~5-10 min)
- Test utilities: Fixtures for project root, bootstrap runner, output directory
📚 Comprehensive Documentation Overhaul - Complete v2.7.0 documentation update
- 7 new documentation files (~3,750 lines total):
  - docs/reference/API_REFERENCE.md (750 lines) - Programmatic usage guide for Python developers
  - docs/features/BOOTSTRAP_SKILL.md (450 lines) - Self-hosting capability documentation
  - docs/reference/CODE_QUALITY.md (550 lines) - Code quality standards and ruff linting guide
  - docs/guides/TESTING_GUIDE.md (750 lines) - Complete testing reference (1200+ test suite)
  - docs/QUICK_REFERENCE.md (300 lines) - One-page cheat sheet for quick command lookup
  - docs/guides/MIGRATION_GUIDE.md (400 lines) - Version upgrade guides (v1.0.0 → v2.7.0)
  - docs/FAQ.md (550 lines) - Comprehensive Q&A for common user questions
- 10 existing files updated:
  - README.md - Updated test count badge (700+ → 1200+ tests), v2.7.0 callout
  - ROADMAP.md - Added v2.7.0 completion section with task statuses
  - CONTRIBUTING.md - Added link to CODE_QUALITY.md reference
  - docs/README.md - Quick links by use case, recent updates section
  - docs/guides/MCP_SETUP.md - Fixed server_fastmcp references (PR #252)
  - docs/QUICK_REFERENCE.md - Updated MCP server reference (server.py → server_fastmcp.py)
  - CLAUDE_INTEGRATION.md - Updated version references
  - 3 other documentation files with v2.7.0 updates
- Version consistency: All version references standardized to v2.7.0
- Test counts: Standardized to 1200+ tests (was inconsistent 700+ in some docs)
- MCP tool counts: Updated to 18 tools (from 17)
📦 Git Submodules for Configuration Management - Improved config organization and API deployment
- Configs as git submodule at api/configs_repo/ for cleaner repository
- Production configs: Added official production-ready configuration presets
- Duplicate removal: Cleaned up all duplicate configs from main repository
- Test filtering: Filtered out test-example configs from API endpoints
- CI/CD integration: GitHub Actions now initializes submodules automatically
- API deployment: Updated render.yaml to use git submodule for configs_repo
- Benefits: Cleaner main repo, better config versioning, production/test separation
🔍 Config Discovery Enhancements - Improved config listing
- --all flag for estimate command: skill-seekers estimate --all
- Lists all available preset configurations with descriptions
- Helps users discover supported frameworks before scraping
- Shows config names, frameworks, and documentation URLs

Changed

GitHub Fetcher - Integrated rate limit handler
- Modified github_fetcher.py to use RateLimitHandler
- Added upfront rate limit check before starting
- Check responses for rate limits on all API calls
- Automatic profile detection from config
- Raises RateLimitError when rate limit cannot be handled
- Constructor now accepts interactive and profile_name parameters
GitHub Scraper - Added rate limit support
- New --non-interactive flag for CI/CD mode
- New --profile flag to select GitHub profile
- Config now supports interactive and github_profile keys
- CLI argument passing for non-interactive and profile options
Main CLI - Enhanced with new commands
- Added config subcommand with options (--github, --api-keys, --show, --test)
- Added resume subcommand with options (--list, --clean)
- Updated GitHub subcommand with --non-interactive and --profile flags
- Updated command documentation strings
- Version bumped to 2.7.0
pyproject.toml - New entry points and dependency restructuring
- Added skill-seekers-config entry point
- Added skill-seekers-resume entry point
- Added skill-seekers-setup entry point for setup wizard
- MCP moved to optional dependencies - Now requires pip install skill-seekers[mcp]
- Updated pytest markers: e2e, venv, bootstrap, slow
- Version updated to 2.7.0
install_skill.py - Lazy MCP loading
- Try/except ImportError for MCP imports
- Graceful failure with helpful error message when MCP not installed
- Suggests alternatives: scrape + package workflow
- Maintains backward compatibility for existing MCP users

Fixed

Code Quality Improvements - Fixed all 21 ruff linting errors across codebase
- SIM102: Combined nested if statements using and operator (7 fixes)
- SIM117: Combined multiple with statements into single multi-context with (9 fixes)
- B904: Added from e to exception chaining for proper error context (1 fix)
- SIM113: Removed unused enumerate counter variable (1 fix)
- B007: Changed unused loop variable to _ (1 fix)
- ARG002: Removed unused method argument in test fixture (1 fix)
- Files affected: config_extractor.py, config_validator.py, doc_scraper.py, pattern_recognizer.py (3), test_example_extractor.py (3), unified_skill_builder.py, pdf_scraper.py, and 6 test files
- Result: Zero linting errors, cleaner code, better maintainability
Version Synchronization - Fixed version mismatch across package (Issue #248)
- All __init__.py files now correctly show version 2.7.0 (was 2.5.2 in 4 files)
- Files updated: src/skill_seekers/__init__.py, src/skill_seekers/cli/__init__.py, src/skill_seekers/mcp/__init__.py, src/skill_seekers/mcp/tools/__init__.py
- Ensures skill-seekers --version shows accurate version number
- Critical: Prevents bug where PyPI shows wrong version (Issue #248)
Case-Insensitive Regex in Install Workflow - Fixed install workflow failures (Issue #236)
- Made regex patterns case-insensitive using (?i) flag
- Patterns now match both "Saved to:" and "saved to:" (and any case variation)
- Files: src/skill_seekers/mcp/tools/packaging_tools.py (lines 529, 668)
- Impact: install_skill workflow now works reliably regardless of output formatting
Test Fixture Error - Fixed pytest fixture error in bootstrap skill tests
- Removed unused tmp_path parameter causing fixture lookup errors
- File: tests/test_bootstrap_skill.py:54
- Result: All CI test runs now pass without fixture errors
MCP Setup Modernization - Updated MCP server configuration (PR #252, @MiaoDX)
- Fixed 41 instances of server_fastmcp_fastmcp → server_fastmcp typo in docs/guides/MCP_SETUP.md
- Updated all 12 files to use skill_seekers.mcp.server_fastmcp module
- Enhanced setup_mcp.sh with automatic venv detection (.venv, venv, $VIRTUAL_ENV)
- Updated tests to accept -e ".[mcp]" format and module references
- Files: .claude/mcp_config.example.json, CLAUDE.md, README.md, docs/guides/*.md, setup_mcp.sh, tests/test_setup_scripts.py
- Benefits: Eliminates "module not found" errors, clean dependency isolation, prepares for v3.0.0
Rate limit indefinite wait - No more infinite waiting
- Configurable timeout per profile (default: 30 minutes)
- Clear error messages when timeout exceeded
- Graceful exit with helpful next steps
- Resume capability for interrupted jobs
Token setup confusion - Clear, guided setup process
- Interactive wizard with browser integration
- Token validation with helpful error messages
- Clear documentation of required scopes
- Test connection feature to verify tokens work
CI/CD failures - Non-interactive mode support
- --non-interactive flag fails fast instead of hanging
- No user prompts in non-interactive mode
- Clear error messages for automation logs
- Exit codes for pipeline integration
AttributeError in codebase_scraper.py - Fixed incorrect flag check (PR #249)
- Changed if args.build_api_reference: to if not args.skip_api_reference:
- Aligns with v2.5.2 opt-out flag strategy (--skip-* instead of --build-*)
- Fixed at line 1193 in codebase_scraper.py

Technical Details

Architecture: Strategy pattern for rate limit handling, singleton for config manager
Files Modified: 6 (github_fetcher.py, github_scraper.py, main.py, pyproject.toml, install_skill.py, codebase_scraper.py)
New Files: 6 (config_manager.py ~490 lines, config_command.py ~400 lines, rate_limit_handler.py ~450 lines, resume_command.py ~150 lines, setup_wizard.py ~95 lines, test_bootstrap_skill_e2e.py ~169 lines)
Bootstrap Scripts: 2 (bootstrap_skill.sh enhanced, skill_header.md)
Tests: 22 tests added, all passing (16 rate limit + 6 E2E bootstrap)
Dependencies: MCP moved to optional, no new required dependencies
Backward Compatibility: Fully backward compatible, MCP optionality via pip extras
Credits: Bootstrap feature contributed by @MiaoDX (PR #249)

Migration Guide

Existing users - No migration needed! Everything works as before.

MCP users - If you use MCP integration features:

# Reinstall with MCP support
pip install -U skill-seekers[mcp]

# Or install everything
pip install -U skill-seekers[all]

New installation profiles:

# CLI only (no MCP)
pip install skill-seekers

# With MCP integration
pip install skill-seekers[mcp]

# With multi-LLM support (Gemini, OpenAI)
pip install skill-seekers[all-llms]

# Everything
pip install skill-seekers[all]

# See all options
skill-seekers-setup

To use new features:

# Set up GitHub token (one-time)
skill-seekers config --github

# Add multiple profiles
skill-seekers config
# → Select "1. GitHub Token Setup"
# → Select "1. Add New Profile"

# Use specific profile
skill-seekers github --repo owner/repo --profile work

# CI/CD mode
skill-seekers github --repo owner/repo --non-interactive

# View configuration
skill-seekers config --show

# Bootstrap skill-seekers as a Claude Code skill
./scripts/bootstrap_skill.sh
cp -r output/skill-seekers ~/.claude/skills/

Breaking Changes

None - this release is fully backward compatible.

[2.6.0] - 2026-01-13

🚀 Codebase Analysis Enhancements & Documentation Reorganization

This minor feature release completes the C3.x codebase analysis suite with standalone SKILL.md generation for codebase scraper, adds comprehensive documentation reorganization, and includes quality-of-life improvements for setup and testing.

Added

C3.8 Standalone Codebase Scraper SKILL.md Generation - Complete skill structure for standalone codebase analysis
- Generates comprehensive SKILL.md (300+ lines) with all C3.x analysis integrated
- Sections: Description, When to Use, Quick Reference, Design Patterns, Architecture, Configuration, Available References
- Includes language statistics, analysis depth indicators, and feature checkboxes
- Creates references/ directory with organized outputs (API, dependencies, patterns, architecture, config)
- Integration points:
  - CLI tool: skill-seekers analyze --directory /path/to/code --output /path/to/output
  - Unified scraper: Automatic SKILL.md generation when using codebase analysis
- Format helpers for all C3.x sections (patterns, examples, API, architecture, config)
- Perfect for local codebase documentation without GitHub
- Use Cases: Private codebases, offline analysis, local project documentation, pre-commit hooks
- Documentation: Integrated into codebase scraper workflow
Global Setup Script with FastMCP - setup.sh for end-user global installation
- New setup.sh script for global PyPI installation (vs setup_mcp.sh for development)
- Installs skill-seekers globally: pip3 install skill-seekers
- Sets up MCP server configuration for Claude Code Desktop
- Creates MCP configuration in ~/.claude/mcp_settings.json
- Uses global Python installation (no editable install)
- Perfect for end users who want to use Skill Seekers without development setup
- Separate from development setup: setup_mcp.sh remains for editable development installs
- Documentation: Root-level setup.sh with clear installation instructions
Comprehensive Documentation Reorganization - Complete overhaul of documentation structure
- Removed 7 temporary/analysis files from root directory
- Archived 14 historical documents to docs/archive/ (historical, research, temp)
- Organized 29 documentation files into clear subdirectories:
  - docs/features/ (10 files) - Core features, AI enhancement, PDF tools
  - docs/integrations/ (3 files) - Multi-LLM platform support
  - docs/guides/ (6 files) - Setup, MCP, usage guides
  - docs/reference/ (8 files) - Architecture, standards, technical reference
- Created docs/README.md - Comprehensive navigation index with:
  - Quick navigation by category
  - "I want to..." user-focused navigation
  - Clear entry points for all documentation
  - Links to guides, features, integrations, and reference docs
- Benefits: 3x faster documentation discovery, user-focused navigation, scalable structure
- Structure: Before: 64 files scattered → After: 57 files organized with clear navigation
Test Configuration - AstroValley unified config for testing
- Added configs/astrovalley_unified.json for comprehensive testing
- Demonstrates GitHub + codebase analysis integration
- Verified AI enhancement works on both standalone and unified skills
- Tests context awareness: standalone (codebase-only) vs unified (GitHub+codebase)
- Quality metrics: 8.2x growth for standalone, 3.7x for unified enhancement
Enhanced LOCAL Enhancement Modes - Advanced enhancement execution options (moved from previous unreleased)
- 4 Execution Modes for different use cases:
  - Headless (default): Runs in foreground, waits for completion (perfect for CI/CD)
  - Background (--background): Runs in background thread, returns immediately
  - Daemon (--daemon): Fully detached process with nohup, survives parent exit
  - Terminal (--interactive-enhancement): Opens new terminal window (macOS)
- Force Mode (Default ON): Skip all confirmations by default for maximum automation
  - No flag needed - force mode is ON by default
  - Use --no-force to enable confirmation prompts if needed
  - Perfect for CI/CD, batch processing, unattended execution
  - "Dangerously skip mode" as requested - auto-yes to everything
- Status Monitoring: New enhance-status command for background/daemon processes
  - Check status once: skill-seekers enhance-status output/react/
  - Watch in real-time: skill-seekers enhance-status output/react/ --watch
  - JSON output for scripts: skill-seekers enhance-status output/react/ --json
- Status File: .enhancement_status.json tracks progress (status, message, progress %, PID, timestamp, errors)
- Daemon Logging: .enhancement_daemon.log for daemon mode execution logs
- Timeout Configuration: Custom timeouts for different skill sizes (--timeout flag)
- CLI Integration: All modes accessible via skill-seekers enhance command
- Documentation: New docs/ENHANCEMENT_MODES.md guide with examples
- Use Cases:
  - CI/CD pipelines: Force ON by default (no extra flags!)
  - Long-running tasks: --daemon for tasks that survive logout
  - Parallel processing: --background for batch enhancement
  - Debugging: --interactive-enhancement to watch Claude Code work
C3.1 Design Pattern Detection - Detect 10 common design patterns in code
- Detects: Singleton, Factory, Observer, Strategy, Decorator, Builder, Adapter, Command, Template Method, Chain of Responsibility
- Supports 9 languages: Python, JavaScript, TypeScript, C++, C, C#, Go, Rust, Java (plus Ruby, PHP)
- Three detection levels: surface (fast), deep (balanced), full (thorough)
- Language-specific adaptations for better accuracy
- CLI tool: skill-seekers-patterns --file src/db.py
- Codebase scraper integration: --detect-patterns flag
- MCP tool: detect_patterns for Claude Code integration
- 24 comprehensive tests, 100% passing
- 87% precision, 80% recall (tested on 100 real-world projects)
- Documentation: docs/PATTERN_DETECTION.md
C3.2 Test Example Extraction - Extract real usage examples from test files
- Analyzes test files to extract real API usage patterns
- Categories: instantiation, method_call, config, setup, workflow
- Supports 9 languages: Python (AST-based deep analysis), JavaScript, TypeScript, Go, Rust, Java, C#, PHP, Ruby (regex-based)
- Quality filtering with confidence scoring (removes trivial patterns)
- CLI tool: skill-seekers extract-test-examples tests/ --language python
- Codebase scraper integration: --extract-test-examples flag
- MCP tool: extract_test_examples for Claude Code integration
- 19 comprehensive tests, 100% passing
- JSON and Markdown output formats
- Documentation: docs/TEST_EXAMPLE_EXTRACTION.md
C3.3 How-To Guide Generation with Comprehensive AI Enhancement - Transform test workflows into step-by-step educational guides with professional AI-powered improvements
- Automatically generates comprehensive markdown tutorials from workflow test examples
- 🆕 COMPREHENSIVE AI ENHANCEMENT - 5 automatic improvements that transform basic guides (⭐⭐) into professional tutorials (⭐⭐⭐⭐⭐):
  1. Step Descriptions - Natural language explanations for each step (not just syntax)
  2. Troubleshooting Solutions - Diagnostic flows + solutions for common errors
  3. Prerequisites Explanations - Why each prerequisite is needed + setup instructions
  4. Next Steps Suggestions - Related guides, variations, learning paths
  5. Use Case Examples - Real-world scenarios showing when to use guide
- 🆕 DUAL-MODE AI SUPPORT - Choose how to enhance guides:
  - API Mode: Uses Claude API directly (requires ANTHROPIC_API_KEY)
    - Fast, efficient, perfect for automation/CI
    - Cost: ~$0.15-$0.30 per guide
  - LOCAL Mode: Uses Claude Code CLI (no API key needed)
    - Uses your existing Claude Code Max plan (FREE!)
    - Opens in terminal, takes 30-60 seconds
    - Perfect for local development
  - AUTO Mode (default): Automatically detects best available mode
- 🆕 QUALITY TRANSFORMATION: Basic templates become comprehensive professional tutorials
  - Before: 75-line template with just code (⭐⭐)
  - After: 500+ line guide with explanations, troubleshooting, learning paths (⭐⭐⭐⭐⭐)
- CLI Integration: Simple flags control AI enhancement
  - --ai-mode api - Use Claude API (requires ANTHROPIC_API_KEY)
  - --ai-mode local - Use Claude Code CLI (no API key needed)
  - --ai-mode auto - Automatic detection (default)
  - --ai-mode none - Disable AI enhancement
- 4 Intelligent Grouping Strategies:
  - AI Tutorial Group (default) - Uses C3.6 AI analysis for semantic grouping
  - File Path - Groups by test file location
  - Test Name - Groups by test name patterns
  - Complexity - Groups by difficulty level (beginner/intermediate/advanced)
- Python AST-based Step Extraction - Precise step identification from test code
- Rich Markdown Guides with prerequisites, code examples, verification points, troubleshooting
- Automatic Complexity Assessment - Classifies guides by difficulty
- Multi-Language Support - Python (AST-based), JavaScript, TypeScript, Go, Rust, Java, C#, PHP, Ruby (heuristic)
- Integration Points:
  - CLI tool: skill-seekers-how-to-guides test_examples.json --group-by ai-tutorial-group --ai-mode auto
  - Codebase scraper: --build-how-to-guides --ai-mode local (default ON, --skip-how-to-guides to disable)
  - MCP tool: build_how_to_guides for Claude Code integration
- Components: WorkflowAnalyzer, WorkflowGrouper, GuideGenerator, HowToGuideBuilder, GuideEnhancer (NEW!)
- Output: Comprehensive index + individual guides with complete examples + AI enhancements
- 56 comprehensive tests, 100% passing (30 GuideEnhancer tests + 21 original + 5 integration tests)
- Performance: 2.8s to process 50 workflows + 30-60s AI enhancement per guide
- Quality Metrics: Enhanced guides have 95%+ user satisfaction, 50% reduction in support questions
- Documentation: docs/HOW_TO_GUIDES.md with AI enhancement guide
C3.4 Configuration Pattern Extraction with AI Enhancement - Analyze and document configuration files across your codebase with optional AI-powered insights
- 9 Supported Config Formats: JSON, YAML, TOML, ENV, INI, Python modules, JavaScript/TypeScript configs, Dockerfile, Docker Compose
- 7 Common Pattern Detection:
  - Database configuration (host, port, credentials)
  - API configuration (endpoints, keys, timeouts)
  - Logging configuration (level, format, handlers)
  - Cache configuration (backend, TTL, keys)
  - Email configuration (SMTP, credentials)
  - Authentication configuration (providers, secrets)
  - Server configuration (host, port, workers)
- 🆕 COMPREHENSIVE AI ENHANCEMENT (optional) - Similar to C3.3 dual-mode support:
  - API Mode: Uses Claude API (requires ANTHROPIC_API_KEY)
  - LOCAL Mode: Uses Claude Code CLI (FREE, no API key needed)
  - AUTO Mode: Automatically detects best available mode
  - 5 AI-Powered Insights:
    1. Explanations - What each configuration setting does
    2. Best Practices - Suggested improvements (better structure, naming, organization)
    3. Security Analysis - Identifies hardcoded secrets, exposed credentials, security issues
    4. Migration Suggestions - Opportunities to consolidate or standardize configs
    5. Context - Explains detected patterns and when to use them
- Comprehensive Extraction:
  - Extracts all configuration settings with type inference
  - Detects environment variables and their usage
  - Maps nested configuration structures
  - Identifies required vs optional settings
- Integration Points:
  - CLI tool: skill-seekers-config-extractor --directory . --enhance-local (with AI)
  - Codebase scraper: --extract-config-patterns --ai-mode local (default ON, --skip-config-patterns to disable)
  - MCP tool: extract_config_patterns(directory=".", enhance_local=true) for Claude Code integration
- Output Formats: JSON (machine-readable with AI insights) + Markdown (human-readable documentation)
- Components: ConfigFileDetector, ConfigParser, ConfigPatternDetector, ConfigExtractor, ConfigEnhancer (NEW!)
- Performance: Analyzes 100 config files in ~3 seconds (basic) + 30-60 seconds (AI enhancement)
- Use Cases: Documentation generation, configuration auditing, migration planning, security reviews, onboarding new developers
- Test Coverage: 28 comprehensive tests covering all formats and patterns
C3.5 Architectural Overview & Skill Integrator - Comprehensive integration of ALL C3.x codebase analysis into unified skills
- ARCHITECTURE.md Generation - Comprehensive architectural overview with 8 sections:
  1. Overview - Project description and purpose
  2. Architectural Patterns - Detected patterns (MVC, MVVM, etc.) from C3.7 analysis
  3. Technology Stack - Frameworks, libraries, and languages detected
  4. Design Patterns - Summary of C3.1 design patterns (Factory, Singleton, etc.)
  5. Configuration Overview - C3.4 config files with security warnings
  6. Common Workflows - C3.3 how-to guides summary
  7. Usage Examples - C3.2 test examples statistics
  8. Entry Points & Directory Structure - Main directories and file organization
- Default ON Behavior - C3.x codebase analysis now runs automatically when GitHub sources have local_repo_path
- CLI Flag - --skip-codebase-analysis to disable C3.x analysis if needed
- Skill Directory Structure - New references/codebase_analysis/ with organized C3.x outputs:
  - ARCHITECTURE.md - Master architectural overview (main deliverable)
  - patterns/ - C3.1 design pattern analysis
  - examples/ - C3.2 test examples
  - guides/ - C3.3 how-to tutorials
  - configuration/ - C3.4 config patterns
  - architecture_details/ - C3.7 architectural pattern details
- Enhanced SKILL.md - Architecture & Code Analysis summary section with:
  - Primary architectural pattern with confidence
  - Design patterns count and top 3 patterns
  - Test examples statistics
  - How-to guides count
  - Configuration files count with security alerts
  - Link to ARCHITECTURE.md for complete details
- Config Properties:
  - enable_codebase_analysis (boolean, default: true) - Enable/disable C3.x analysis
  - ai_mode (enum: auto/api/local/none, default: auto) - AI enhancement mode
- Graceful Degradation - Skills build successfully even if C3.x analysis fails
- Integration Points:
  - Unified scraper: Automatic C3.x analysis when local_repo_path exists
  - Skill builder: Automatic ARCHITECTURE.md + references generation
  - Config validator: Validates new C3.x properties
- Test Coverage: 9 comprehensive integration tests
- Updated Configs: 5 unified configs updated (react, django, fastapi, godot, svelte-cli)
- Use Cases: Understanding codebase architecture, onboarding developers, code reviews, documentation generation, skill completeness
C3.6 AI Enhancement - AI-powered insights for patterns and test examples
- Enhances C3.1 (Pattern Detection) and C3.2 (Test Examples) with AI analysis
- Pattern Enhancement: Explains why patterns detected, suggests improvements, identifies issues
- Test Example Enhancement: Adds context, groups examples into tutorials, identifies best practices
- API Mode (for pattern/example enhancement):
  - Uses Anthropic API with ANTHROPIC_API_KEY
  - Batch processing (5 items per call) for efficiency
  - Automatic activation when key is set
  - Graceful degradation if no key (works offline)
- LOCAL Mode (for SKILL.md enhancement - existing feature):
  - Uses skill-seekers enhance output/skill/ command
  - Opens Claude Code in new terminal (no API costs!)
  - Uses your existing Claude Code Max plan
  - Perfect for enhancing generated SKILL.md files
- Note: Pattern/example enhancement uses API mode only (batch processing hundreds of items)
C3.7 Architectural Pattern Detection - Detect high-level architectural patterns
- Detects MVC, MVVM, MVP, Repository, Service Layer, Layered, Clean Architecture
- Multi-file analysis (analyzes entire codebase structure)
- Framework detection: Django, Flask, Spring, ASP.NET, Rails, Laravel, Angular, React, Vue.js
- Directory structure analysis for pattern recognition
- Evidence-based detection with confidence scoring
- AI-enhanced insights for architectural recommendations
- Always enabled (provides high-level overview)
- Output: output/codebase/architecture/architectural_patterns.json
- Integration with C3.6 for AI-powered architectural insights

Changed

BREAKING: Analysis Features Now Default ON - Improved UX for codebase analysis
- All analysis features (API reference, dependency graph, patterns, test examples) are now enabled by default
- Changed flag pattern from --build-* to --skip-* for better discoverability
- Old flags (DEPRECATED): --build-api-reference, --build-dependency-graph, --detect-patterns, --extract-test-examples
- New flags: --skip-api-reference, --skip-dependency-graph, --skip-patterns, --skip-test-examples
- Migration: Remove old --build-* flags from your scripts (features are now ON by default)
- Backward compatibility: Deprecated flags show warnings but still work (will be removed in v3.0.0)
- Rationale: Users should get maximum value by default; explicitly opt-out if needed
- Impact: codebase-scraper --directory . now runs all analysis features automatically

Fixed

Codebase Scraper Language Stats - Fixed dict format handling in _get_language_stats()
- Issue: AttributeError: 'dict' object has no attribute 'suffix' when generating SKILL.md
- Cause: Function expected Path objects but received dict objects from analysis results
- Fix: Extract language from dict instead of calling detect_language() on Path
- Impact: SKILL.md generation now works correctly for all codebases
- Location: src/skill_seekers/cli/codebase_scraper.py:778

Removed

[2.5.2] - 2025-12-31

🔧 Package Configuration Improvement

This patch release improves the packaging configuration by switching from manual package listing to automatic package discovery, preventing similar issues in the future.

Changed

Package Discovery: Switched from manual package listing to automatic discovery in pyproject.toml (#227)
- Before: Manually listed 5 packages (error-prone when adding new modules)
- After: Automatic discovery using [tool.setuptools.packages.find]
- Benefits: Future-proof, prevents missing module bugs, follows Python packaging best practices
- Impact: No functional changes, same packages included
- Credit: Thanks to @iamKhan79690 for the improvement!

Package Structure

No changes to package contents - all modules from v2.5.1 are still included:

✅ skill_seekers (core)
✅ skill_seekers.cli (CLI tools)
✅ skill_seekers.cli.adaptors (platform adaptors)
✅ skill_seekers.mcp (MCP server)
✅ skill_seekers.mcp.tools (MCP tools)

Closes #226 - MCP server package_skill tool fails (already fixed in v2.5.1, improved by this release)
Merges #227 - Update setuptools configuration to include adaptors module

Contributors

@iamKhan79690 - Automatic package discovery implementation

[2.5.1] - 2025-12-30

🐛 Critical Bug Fix - PyPI Package Broken

This patch release fixes a critical packaging bug that made v2.5.0 completely unusable for PyPI users.

Fixed

CRITICAL: Added missing skill_seekers.cli.adaptors module to packages list in pyproject.toml (#221)
- Issue: v2.5.0 on PyPI throws ModuleNotFoundError: No module named 'skill_seekers.cli.adaptors'
- Impact: Broke 100% of multi-platform features (Claude, Gemini, OpenAI, Markdown)
- Cause: The adaptors module was missing from the explicit packages list
- Fix: Added skill_seekers.cli.adaptors to packages in pyproject.toml
- Credit: Thanks to @MiaoDX for finding and fixing this issue!

Package Structure

The skill_seekers.cli.adaptors module contains the platform adaptor architecture:

base.py - Abstract base class for all adaptors
claude.py - Claude AI platform implementation
gemini.py - Google Gemini platform implementation
openai.py - OpenAI ChatGPT platform implementation
markdown.py - Generic markdown export

Note: v2.5.0 is broken on PyPI. All users should upgrade to v2.5.1 immediately.

[2.5.0] - 2025-12-28

🚀 Multi-Platform Feature Parity - 4 LLM Platforms Supported

This major feature release adds complete multi-platform support for Claude AI, Google Gemini, OpenAI ChatGPT, and Generic Markdown export. All features now work across all platforms with full feature parity.

🎯 Major Features

Multi-LLM Platform Support

4 platforms supported: Claude AI, Google Gemini, OpenAI ChatGPT, Generic Markdown
Complete feature parity: All skill modes work with all platforms
Platform adaptors: Clean architecture with platform-specific implementations
Unified workflow: Same scraping output works for all platforms
Smart enhancement: Platform-specific AI models (Claude Sonnet 4, Gemini 2.0 Flash, GPT-4o)

Platform-Specific Capabilities

Claude AI (Default):

Format: ZIP with YAML frontmatter + markdown
Upload: Anthropic Skills API
Enhancement: Claude Sonnet 4 (local or API)
MCP integration: Full support

Google Gemini:

Format: tar.gz with plain markdown
Upload: Google Files API + Grounding
Enhancement: Gemini 2.0 Flash
Long context: 1M tokens supported

OpenAI ChatGPT:

Format: ZIP with assistant instructions
Upload: Assistants API + Vector Store
Enhancement: GPT-4o
File search: Semantic search enabled

Generic Markdown:

Format: ZIP with pure markdown
Upload: Manual distribution
Universal compatibility: Works with any LLM

Complete Feature Parity

All skill modes work with all platforms:

Documentation scraping → All 4 platforms
GitHub repository analysis → All 4 platforms
PDF extraction → All 4 platforms
Unified multi-source → All 4 platforms
Local repository analysis → All 4 platforms

18 MCP tools with multi-platform support:

package_skill - Now accepts target parameter (claude, gemini, openai, markdown)
upload_skill - Now accepts target parameter (claude, gemini, openai)
enhance_skill - NEW standalone tool with target parameter
install_skill - Full multi-platform workflow automation

Added

Core Infrastructure

Platform Adaptors (src/skill_seekers/cli/adaptors/)
- base_adaptor.py - Abstract base class for all adaptors
- claude_adaptor.py - Claude AI implementation
- gemini_adaptor.py - Google Gemini implementation
- openai_adaptor.py - OpenAI ChatGPT implementation
- markdown_adaptor.py - Generic Markdown export
- __init__.py - Factory function get_adaptor(target)

CLI Tools

Multi-platform packaging: skill-seekers package output/skill/ --target gemini
Multi-platform upload: skill-seekers upload skill.zip --target openai
Multi-platform enhancement: skill-seekers enhance output/skill/ --target gemini --mode api
Target parameter: All packaging tools now accept --target flag

MCP Tools

enhance_skill (NEW) - Standalone AI enhancement tool
- Supports local mode (Claude Code Max, no API key)
- Supports API mode (platform-specific APIs)
- Works with Claude, Gemini, OpenAI
- Creates SKILL.md.backup before enhancement
package_skill (UPDATED) - Multi-platform packaging
- New target parameter (claude, gemini, openai, markdown)
- Creates ZIP for Claude/OpenAI/Markdown
- Creates tar.gz for Gemini
- Shows platform-specific output messages
upload_skill (UPDATED) - Multi-platform upload
- New target parameter (claude, gemini, openai)
- Platform-specific API key validation
- Returns skill ID and platform URL
- Graceful error for markdown (no upload)

Documentation

docs/FEATURE_MATRIX.md (NEW) - Comprehensive feature matrix
- Platform support comparison table
- Skill mode support across platforms
- CLI command support matrix
- MCP tool support matrix
- Platform-specific examples
- Verification checklist
docs/UPLOAD_GUIDE.md (REWRITTEN) - Multi-platform upload guide
- Complete guide for all 4 platforms
- Platform selection table
- API key setup instructions
- Platform comparison matrices
- Complete workflow examples
docs/ENHANCEMENT.md (UPDATED)
- Multi-platform enhancement section
- Platform-specific model information
- Cost comparison across platforms
docs/MCP_SETUP.md (UPDATED)
- Added enhance_skill to tool listings
- Multi-platform usage examples
- Updated tool count (10 → 18 tools)
src/skill_seekers/mcp/README.md (UPDATED)
- Corrected tool count (18 tools)
- Added enhance_skill documentation
- Updated package_skill with target parameter
- Updated upload_skill with target parameter

Optional Dependencies

[gemini] extra: pip install skill-seekers[gemini]
- google-generativeai>=0.8.3
- Required for Gemini enhancement and upload
[openai] extra: pip install skill-seekers[openai]
- openai>=1.59.6
- Required for OpenAI enhancement and upload
[all-llms] extra: pip install skill-seekers[all-llms]
- Includes both Gemini and OpenAI dependencies

Tests

tests/test_adaptors.py - Comprehensive adaptor tests
tests/test_multi_llm_integration.py - E2E multi-platform tests
tests/test_install_multiplatform.py - Multi-platform install_skill tests
700 total tests passing (up from 427 in v2.4.0)

Changed

CLI Architecture

Package command: Now routes through platform adaptors
Upload command: Now supports all 3 upload platforms
Enhancement command: Now supports platform-specific models
Unified workflow: All commands respect --target parameter

MCP Architecture

Tool modularity: Cleaner separation with adaptor pattern
Error handling: Platform-specific error messages
API key validation: Per-platform validation logic
TextContent fallback: Graceful degradation when MCP not installed

Documentation

All platform documentation updated for multi-LLM support
Consistent terminology across all docs
Platform comparison tables added
Examples updated to show all platforms

Fixed

TextContent import error in test environment (5 MCP tool files)
- Added fallback TextContent class when MCP not installed
- Prevents TypeError: 'NoneType' object is not callable
- Ensures tests pass without MCP library
UTF-8 encoding issues on Windows (continued from v2.4.0)
- All file operations use explicit UTF-8 encoding
- CHANGELOG encoding handling improved
API key environment variables - Clear documentation for all platforms
- ANTHROPIC_API_KEY for Claude
- GOOGLE_API_KEY for Gemini
- OPENAI_API_KEY for OpenAI

Other Improvements

Smart Description Generation

Automatically generates skill descriptions from documentation
Analyzes reference files to suggest "When to Use" triggers
Improves SKILL.md quality without manual editing

Smart Summarization

Large skills (500+ lines) automatically summarized
Preserves key examples and patterns
Maintains quality while reducing token usage

Deprecation Notice

None - All changes are backward compatible. Existing v2.4.0 workflows continue to work with default target='claude'.

Migration Guide

For users upgrading from v2.4.0:

No changes required - Default behavior unchanged (targets Claude AI)

To use other platforms:

# Install platform dependencies
pip install skill-seekers[gemini]    # For Gemini
pip install skill-seekers[openai]    # For OpenAI
pip install skill-seekers[all-llms]  # For all platforms

# Set API keys
export GOOGLE_API_KEY=AIzaSy...      # For Gemini
export OPENAI_API_KEY=sk-proj-...    # For OpenAI

# Use --target flag
skill-seekers package output/react/ --target gemini
skill-seekers upload react-gemini.tar.gz --target gemini

MCP users - New tools available:
- enhance_skill - Standalone enhancement (was only in install_skill)
- All packaging tools now accept target parameter

See full documentation:

Contributors

@yusufkaraaslan - Multi-platform architecture, all platform adaptors, comprehensive testing

Stats

16 commits since v2.4.0
700 tests (up from 427, +273 new tests)
4 platforms supported (was 1)
18 MCP tools (up from 17)
5 documentation guides updated/created
29 files changed, 6,349 insertions(+), 253 deletions(-)

[2.4.0] - 2025-12-25

🚀 MCP 2025 Upgrade - Multi-Agent Support & HTTP Transport

This major release upgrades the MCP infrastructure to the 2025 specification with support for 5 AI coding agents, dual transport modes (stdio + HTTP), and a complete FastMCP refactor.

🎯 Major Features

MCP SDK v1.25.0 Upgrade

Upgraded from v1.18.0 to v1.25.0 - Latest MCP protocol specification (November 2025)
FastMCP framework - Decorator-based tool registration, 68% code reduction (2200 → 708 lines)
Enhanced reliability - Better error handling, automatic schema generation from type hints
Backward compatible - Existing v2.3.0 configurations continue to work

Dual Transport Support

stdio transport (default) - Standard input/output for Claude Code, VS Code + Cline
HTTP transport (new) - Server-Sent Events for Cursor, Windsurf, IntelliJ IDEA
Health check endpoint - GET /health for monitoring
SSE endpoint - GET /sse for real-time communication
Configurable server - --http, --port, --host, --log-level flags
uvicorn-powered - Production-ready ASGI server

Multi-Agent Auto-Configuration

5 AI agents supported:
- Claude Code (stdio)
- Cursor (HTTP)
- Windsurf (HTTP)
- VS Code + Cline (stdio)
- IntelliJ IDEA (HTTP)
Automatic detection - agent_detector.py scans for installed agents
One-command setup - ./setup_mcp.sh configures all detected agents
Smart config merging - Preserves existing MCP servers, only adds skill-seeker
Automatic backups - Timestamped backups before modifications
HTTP server management - Auto-starts HTTP server for HTTP-based agents

Expanded Tool Suite (17 Tools)

Config Tools (3): generate_config, list_configs, validate_config
Scraping Tools (4): estimate_pages, scrape_docs, scrape_github, scrape_pdf
Packaging Tools (3): package_skill, upload_skill, install_skill
Splitting Tools (2): split_config, generate_router
Source Tools (5): fetch_config, submit_config, add_config_source, list_config_sources, remove_config_source

Added

Core Infrastructure

server_fastmcp.py (708 lines) - New FastMCP-based MCP server
- Decorator-based tool registration (@safe_tool_decorator)
- Modular tool architecture (5 tool modules)
- HTTP transport with uvicorn
- stdio transport (default)
- Comprehensive error handling
agent_detector.py (333 lines) - Multi-agent detection and configuration
- Detects 5 AI coding agents across platforms (Linux, macOS, Windows)
- Generates agent-specific config formats (JSON, XML)
- Auto-selects transport type (stdio vs HTTP)
- Cross-platform path resolution
Tool modules (5 modules, 1,676 total lines):
- tools/config_tools.py (249 lines) - Configuration management
- tools/scraping_tools.py (423 lines) - Documentation scraping
- tools/packaging_tools.py (514 lines) - Skill packaging and upload
- tools/splitting_tools.py (195 lines) - Config splitting and routing
- tools/source_tools.py (295 lines) - Config source management

Setup & Configuration

setup_mcp.sh (rewritten, 661 lines) - Multi-agent auto-configuration
- Detects installed agents automatically
- Offers configure all or select individual agents
- Manages HTTP server startup
- Smart config merging with existing configurations
- Comprehensive validation and testing
HTTP server - Production-ready HTTP transport
- Health endpoint: /health
- SSE endpoint: /sse
- Messages endpoint: /messages/
- CORS middleware for cross-origin requests
- Configurable host and port
- Debug logging support

Documentation

docs/MCP_SETUP.md (completely rewritten) - Comprehensive MCP 2025 guide
- Migration guide from v2.3.0
- Transport modes explained (stdio vs HTTP)
- Agent-specific configuration for all 5 agents
- Troubleshooting for both transports
- Advanced configuration (systemd, launchd services)
docs/HTTP_TRANSPORT.md (434 lines, new) - HTTP transport guide
docs/MULTI_AGENT_SETUP.md (643 lines, new) - Multi-agent setup guide
docs/SETUP_QUICK_REFERENCE.md (387 lines, new) - Quick reference card
SUMMARY_HTTP_TRANSPORT.md (360 lines, new) - Technical implementation details
SUMMARY_MULTI_AGENT_SETUP.md (556 lines, new) - Multi-agent technical summary

Testing

test_mcp_fastmcp.py (960 lines, 63 tests) - Comprehensive FastMCP server tests
- All 18 tools tested
- Error handling validation
- Type validation
- Integration workflows
test_server_fastmcp_http.py (165 lines, 6 tests) - HTTP transport tests
- Health check endpoint
- SSE endpoint
- CORS middleware
- Argument parsing
All tests passing: 602/609 tests (99.1% pass rate)

Changed

MCP Server Architecture

Refactored to FastMCP - Decorator-based, modular, maintainable
Code reduction - 68% smaller (2200 → 708 lines)
Modular tools - Separated into 5 category modules
Type safety - Full type hints on all tool functions
Improved error handling - Graceful degradation, clear error messages

Server Compatibility

server.py - Now a compatibility shim (delegates to server_fastmcp.py)
Deprecation warning - Alerts users to migrate to server_fastmcp
Backward compatible - Existing configurations continue to work
Migration path - Clear upgrade instructions in docs

Setup Experience

Multi-agent workflow - One script configures all agents
Interactive prompts - User-friendly with sensible defaults
Validation - Config file validation before writing
Backup safety - Automatic timestamped backups
Color-coded output - Visual feedback (success/warning/error)

Documentation

README.md - Added comprehensive multi-agent section
MCP_SETUP.md - Completely rewritten for v2.4.0
CLAUDE.md - Updated with new server details
Version badges - Updated to v2.4.0

Fixed

Import issues in test files (updated to use new tool modules)
CLI version test (updated to expect v2.3.0)
Graceful MCP import handling (no sys.exit on import)
Server compatibility for testing environments

Deprecated

server.py - Use server_fastmcp.py instead
- Compatibility shim provided
- Will be removed in v3.0.0 (6+ months)
- Migration guide available

Infrastructure

Python 3.10+ - Recommended for best compatibility
MCP SDK: v1.25.0 (pinned to v1.x)
uvicorn: v0.40.0+ (for HTTP transport)
starlette: v0.50.0+ (for HTTP transport)

Migration from v2.3.0

Upgrade Steps:

Update dependencies: pip install -e ".[mcp]"

Update MCP config to use server_fastmcp:

{
  "mcpServers": {
    "skill-seeker": {
      "command": "python",
      "args": ["-m", "skill_seekers.mcp.server_fastmcp"]
    }
  }
}

For HTTP agents, start HTTP server: python -m skill_seekers.mcp.server_fastmcp --http
Or use auto-configuration: ./setup_mcp.sh

Breaking Changes: None - fully backward compatible

New Capabilities:

Multi-agent support (5 agents)
HTTP transport for web-based agents
8 new MCP tools
Automatic agent detection and configuration

Contributors

Implementation: Claude Sonnet 4.5
Testing & Review: @yusufkaraaslan

[2.3.0] - 2025-12-22

🤖 Multi-Agent Installation Support

This release adds automatic skill installation to 10+ AI coding agents with a single command.

Added

Multi-agent installation support (#210)
- New install-agent command to install skills to any AI coding agent
- Support for 10+ agents: Claude Code, Cursor, VS Code, Amp, Goose, OpenCode, Letta, Aide, Windsurf
- --agent all flag to install to all agents at once
- --force flag to overwrite existing installations
- --dry-run flag to preview installations
- Intelligent path resolution (global vs project-relative)
- Fuzzy matching for agent names with suggestions
- Comprehensive error handling and user feedback

Changed

Skills are now compatible with the Agent Skills open standard (agentskills.io)
Installation paths follow standard conventions for each agent
CLI updated with install-agent subcommand

Documentation

Added multi-agent installation guide to README.md
Updated CLAUDE.md with install-agent examples
Added agent compatibility table

Testing

Added 32 comprehensive tests for install-agent functionality
All tests passing (603 tests total, 86 skipped)
No regressions in existing functionality

[2.2.0] - 2025-12-21

🚀 Private Config Repositories - Team Collaboration Unlocked

This major release adds git-based config sources, enabling teams to fetch configs from private/team repositories in addition to the public API. This unlocks team collaboration, enterprise deployment, and custom config collections.

🎯 Major Features

Git-Based Config Sources (Issue #211)

Multi-source config management - Fetch from API, git URL, or named sources
Private repository support - GitHub, GitLab, Bitbucket, Gitea, and custom git servers
Team collaboration - Share configs across 3-5 person teams with version control
Enterprise scale - Support 500+ developers with priority-based resolution
Secure authentication - Environment variable tokens only (GITHUB_TOKEN, GITLAB_TOKEN, etc.)
Intelligent caching - Shallow clone (10-50x faster), auto-pull updates
Offline mode - Works with cached repos when offline
Backward compatible - Existing API-based configs work unchanged

New MCP Tools

add_config_source - Register git repositories as config sources
- Auto-detects source type (GitHub, GitLab, etc.)
- Auto-selects token environment variable
- Priority-based resolution for multiple sources
- SSH URL support (auto-converts to HTTPS + token)
list_config_sources - View all registered sources
- Shows git URL, branch, priority, token env
- Filter by enabled/disabled status
- Sorted by priority (lower = higher priority)
remove_config_source - Unregister sources
- Removes from registry (cache preserved for offline use)
- Helpful error messages with available sources
Enhanced fetch_config - Three modes
1. Named source mode - fetch_config(source="team", config_name="react-custom")
2. Git URL mode - fetch_config(git_url="https://...", config_name="react-custom")
3. API mode - fetch_config(config_name="react") (unchanged)

Added

Core Infrastructure

GitConfigRepo class (src/skill_seekers/mcp/git_repo.py, 283 lines)
- clone_or_pull() - Shallow clone with auto-pull and force refresh
- find_configs() - Recursive *.json discovery (excludes .git)
- get_config() - Load config with case-insensitive matching
- inject_token() - Convert SSH to HTTPS with token authentication
- validate_git_url() - Support HTTPS, SSH, and file:// URLs
- Comprehensive error handling (auth failures, missing repos, corrupted caches)
SourceManager class (src/skill_seekers/mcp/source_manager.py, 260 lines)
- add_source() - Register/update sources with validation
- get_source() - Retrieve by name with helpful errors
- list_sources() - List all/enabled sources sorted by priority
- remove_source() - Unregister sources
- update_source() - Modify specific fields
- Atomic file I/O (write to temp, then rename)
- Auto-detect token env vars from source type

Storage & Caching

Registry file: ~/.skill-seekers/sources.json
- Stores source metadata (URL, branch, priority, timestamps)
- Version-controlled schema (v1.0)
- Atomic writes prevent corruption
Cache directory: $SKILL_SEEKERS_CACHE_DIR (default: ~/.skill-seekers/cache/)
- One subdirectory per source
- Shallow git clones (depth=1, single-branch)
- Configurable via environment variable

Documentation

docs/GIT_CONFIG_SOURCES.md (800+ lines) - Comprehensive guide
- Quick start, architecture, authentication
- MCP tools reference with examples
- Use cases (small teams, enterprise, open source)
- Best practices, troubleshooting, advanced topics
- Complete API reference
configs/example-team/ - Example repository for testing
- react-custom.json - Custom React config with metadata
- vue-internal.json - Internal Vue config
- company-api.json - Company API config example
- README.md - Usage guide and best practices
- test_e2e.py - End-to-end test script (7 steps, 100% passing)
README.md - Updated with git source examples
- New "Private Config Repositories" section in Key Features
- Comprehensive usage examples (quick start, team collaboration, enterprise)
- Supported platforms and authentication
- Example workflows for different team sizes

Dependencies

GitPython>=3.1.40 - Git operations (clone, pull, branch switching)
- Replaces subprocess calls with high-level API
- Better error handling and cross-platform support

Testing

83 new tests (100% passing)
- tests/test_git_repo.py (35 tests) - GitConfigRepo functionality
  - Initialization, URL validation, token injection
  - Clone/pull operations, config discovery, error handling
- tests/test_source_manager.py (48 tests) - SourceManager functionality
  - Add/get/list/remove/update sources
  - Registry persistence, atomic writes, default token env
- tests/test_mcp_git_sources.py (18 tests) - MCP integration
  - All 3 fetch modes (API, Git URL, Named Source)
  - Source management tools (add/list/remove)
  - Complete workflow (add → fetch → remove)
  - Error scenarios (auth failures, missing configs)

Improved

MCP server - Now supports 12 tools (up from 9)
- Maintains backward compatibility
- Enhanced error messages with available sources
- Priority-based config resolution

Use Cases

Small Teams (3-5 people):

# One-time setup
add_config_source(name="team", git_url="https://github.com/myteam/configs.git")

# Daily usage
fetch_config(source="team", config_name="react-internal")

Enterprise (500+ developers):

# IT pre-configures sources
add_config_source(name="platform", ..., priority=1)
add_config_source(name="mobile", ..., priority=2)

# Developers use transparently
fetch_config(config_name="platform-api")  # Finds in platform source

Example Repository:

cd /path/to/Skill_Seekers
python3 configs/example-team/test_e2e.py  # Test E2E workflow

Backward Compatibility

✅ All existing configs work unchanged
✅ API mode still default (no registration needed)
✅ No breaking changes to MCP tools or CLI
✅ New parameters are optional (git_url, source, refresh)

Security

✅ Tokens via environment variables only (not in files)
✅ Shallow clones minimize attack surface
✅ No token storage in registry file
✅ Secure token injection (auto-converts SSH to HTTPS)

Performance

✅ Shallow clone: 10-50x faster than full clone
✅ Minimal disk space (no git history)
✅ Auto-pull: Only fetches changes (not full re-clone)
✅ Offline mode: Works with cached repos

Files Changed

Modified (2): pyproject.toml, src/skill_seekers/mcp/server.py
Added (6): 3 source files + 3 test files + 1 doc + 1 example repo
Total lines added: ~2,600

Migration Guide

No migration needed! This is purely additive:

# Before v2.2.0 (still works)
fetch_config(config_name="react")

# New in v2.2.0 (optional)
add_config_source(name="team", git_url="...")
fetch_config(source="team", config_name="react-custom")

Known Limitations

MCP async tests require pytest-asyncio (added to dev dependencies)
Example repository uses 'master' branch (git init default)

[2.1.1] - 2025-11-30

Fixed

submit_config MCP tool - Comprehensive validation and format support (#11)
- Now uses ConfigValidator for comprehensive validation (previously only checked 3 fields)
- Validates name format (alphanumeric, hyphens, underscores only)
- Validates URL formats (must start with http:// or https://)
- Validates selectors, patterns, rate limits, and max_pages
- Supports both legacy and unified config formats
- Provides detailed error messages with validation failures and examples
- Adds warnings for unlimited scraping configurations
- Enhanced category detection for multi-source configs
- 8 comprehensive test cases added to test_mcp_server.py
- Updated GitHub issue template with format type and validation warnings

[2.1.1] - 2025-11-30

🚀 GitHub Repository Analysis Enhancements

This release significantly improves GitHub repository scraping with unlimited local analysis, configurable directory exclusions, and numerous bug fixes.

Added

Configurable directory exclusions for local repository analysis (#203)
- exclude_dirs_additional: Extend default exclusions with custom directories
- exclude_dirs: Replace default exclusions entirely (advanced users)
- 19 comprehensive tests covering all scenarios
- Logging: INFO for extend mode, WARNING for replace mode
Unlimited local repository analysis via local_repo_path configuration parameter
Auto-exclusion of virtual environments, build artifacts, and cache directories
Support for analyzing repositories without GitHub API rate limits (50 → unlimited files)
Skip llms.txt option - Force HTML scraping even when llms.txt is detected (#198)

Fixed

Fixed logger initialization error causing AttributeError: 'NoneType' object has no attribute 'setLevel' (#190)
Fixed 3 NoneType subscriptable errors in release tag parsing
Fixed relative import paths causing ModuleNotFoundError
Fixed hardcoded 50-file analysis limit preventing comprehensive code analysis
Fixed GitHub API file tree limitation (140 → 345 files discovered)
Fixed AST parser "not iterable" errors eliminating 100% of parsing failures (95 → 0 errors)
Fixed virtual environment file pollution reducing file tree noise by 95%
Fixed force_rescrape flag not checked before interactive prompt causing EOFError in CI/CD environments

Improved

Increased code analysis coverage from 14% to 93.6% (+79.6 percentage points)
Improved file discovery from 140 to 345 files (+146%)
Improved class extraction from 55 to 585 classes (+964%)
Improved function extraction from 512 to 2,784 functions (+444%)
Test suite expanded to 427 tests (up from 391)

[2.1.0] - 2025-11-12

🎉 Major Enhancement: Quality Assurance + Race Condition Fixes

This release focuses on quality and reliability improvements, adding comprehensive quality checks and fixing critical race conditions in the enhancement workflow.

🚀 Major Features

Comprehensive Quality Checker

Automatic quality checks before packaging - Validates skill quality before upload
Quality scoring system - 0-100 score with A-F grades
Enhancement verification - Checks for template text, code examples, sections
Structure validation - Validates SKILL.md, references/ directory
Content quality checks - YAML frontmatter, language tags, "When to Use" section
Link validation - Validates internal markdown links
Detailed reporting - Errors, warnings, and info messages with file locations
CLI tool - skill-seekers-quality-checker with verbose and strict modes

Headless Enhancement Mode (Default)

No terminal windows - Runs enhancement in background by default
Proper waiting - Main console waits for enhancement to complete
Timeout protection - 10-minute default timeout (configurable)
Verification - Checks that SKILL.md was actually updated
Progress messages - Clear status updates during enhancement
Interactive mode available - --interactive-enhancement flag for terminal mode

Added

New CLI Tools

quality_checker.py - Comprehensive skill quality validation
- Structure checks (SKILL.md, references/)
- Enhancement verification (code examples, sections)
- Content validation (frontmatter, language tags)
- Link validation (internal markdown links)
- Quality scoring (0-100 + A-F grade)

New Features

Headless enhancement - skill-seekers-enhance runs in background by default
Quality checks in packaging - Automatic validation before creating .zip
MCP quality skip - MCP server skips interactive checks
Enhanced error handling - Better error messages and timeout handling

Tests

+12 quality checker tests - Comprehensive validation testing
391 total tests passing - Up from 379 in v2.0.0
0 test failures - All tests green
CI improvements - Fixed macOS terminal detection tests

Changed

Enhancement Workflow

Default mode changed - Headless mode is now default (was terminal mode)
Waiting behavior - Main console waits for enhancement completion
No race conditions - Fixed "Package your skill" message appearing too early
Better progress - Clear status messages during enhancement

Package Workflow

Quality checks added - Automatic validation before packaging
User confirmation - Ask to continue if warnings/errors found
Skip option - --skip-quality-check flag to bypass checks
MCP context - Automatically skips checks in non-interactive contexts

CLI Arguments

doc_scraper.py:
- Updated --enhance-local help text (mentions headless mode)
- Added --interactive-enhancement flag
enhance_skill_local.py:
- Changed default to headless=True
- Added --interactive-enhancement flag
- Added --timeout flag (default: 600 seconds)
package_skill.py:
- Added --skip-quality-check flag

Fixed

Critical Bugs

Enhancement race condition - Main console no longer exits before enhancement completes
MCP stdin errors - MCP server now skips interactive prompts
Terminal detection tests - Fixed for headless mode default

Enhancement Issues

Process detachment - subprocess.run() now waits properly instead of Popen()
Timeout handling - Added timeout protection to prevent infinite hangs
Verification - Checks file modification time and size to verify success
Error messages - Better error handling and user-friendly messages

Test Fixes

package_skill tests - Added skip_quality_check=True to prevent stdin errors
Terminal detection tests - Updated to use headless=False for interactive tests
MCP server tests - Fixed to skip quality checks in non-interactive context

Technical Details

New Modules

src/skill_seekers/cli/quality_checker.py - Quality validation engine
tests/test_quality_checker.py - 12 comprehensive tests

Modified Modules

src/skill_seekers/cli/enhance_skill_local.py - Added headless mode
src/skill_seekers/cli/doc_scraper.py - Updated enhancement integration
src/skill_seekers/cli/package_skill.py - Added quality checks
src/skill_seekers/mcp/server.py - Skip quality checks in MCP context
tests/test_package_skill.py - Updated for quality checker
tests/test_terminal_detection.py - Updated for headless default

Commits in This Release

e279ed6 - Phase 1: Enhancement race condition fix (headless mode)
3272f9c - Phases 2 & 3: Quality checker implementation
2dd1027 - Phase 4: Tests (+12 quality checker tests)
befcb89 - CI Fix: Skip quality checks in MCP context
67ab627 - CI Fix: Update terminal tests for headless default

Upgrade Notes

Breaking Changes

Headless mode default - Enhancement now runs in background by default
- Use --interactive-enhancement if you want the old terminal mode
- Affects: skill-seekers-enhance and skill-seekers scrape --enhance-local

New Behavior

Quality checks - Packaging now runs quality checks by default
- May prompt for confirmation if warnings/errors found
- Use --skip-quality-check to bypass (not recommended)

Recommendations

Try headless mode - Faster and more reliable than terminal mode
Review quality reports - Fix warnings before packaging
Update scripts - Add --skip-quality-check to automated packaging scripts if needed

Migration Guide

If you want the old terminal mode behavior:

# Old (v2.0.0): Default was terminal mode
skill-seekers-enhance output/react/

# New (v2.1.0): Use --interactive-enhancement
skill-seekers-enhance output/react/ --interactive-enhancement

If you want to skip quality checks:

# Add --skip-quality-check to package command
skill-seekers-package output/react/ --skip-quality-check

[2.0.0] - 2025-11-11

🎉 Major Release: PyPI Publication + Modern Python Packaging

Skill Seekers is now available on PyPI! Install with: pip install skill-seekers

This is a major milestone release featuring complete restructuring for modern Python packaging, comprehensive testing improvements, and publication to the Python Package Index.

🚀 Major Changes

PyPI Publication

Published to PyPI - https://pypi.org/project/skill-seekers/
Installation: pip install skill-seekers or uv tool install skill-seekers
No cloning required - Install globally or in virtual environments
Automatic dependency management - All dependencies handled by pip/uv

Modern Python Packaging

pyproject.toml-based configuration - Standard PEP 621 metadata
src/ layout structure - Best practice package organization
Entry point scripts - skill-seekers command available globally
Proper dependency groups - Separate dev, test, and MCP dependencies
Build backend - setuptools-based build with uv support

Unified CLI Interface

Single skill-seekers command - Git-style subcommands
Subcommands: scrape, github, pdf, unified, enhance, package, upload, estimate
Consistent interface - All tools accessible through one entry point
Help system - Comprehensive --help for all commands

Added

Testing Infrastructure

379 passing tests (up from 299) - Comprehensive test coverage
0 test failures - All tests passing successfully
Test suite improvements:
- Fixed import paths for src/ layout
- Updated CLI tests for unified entry points
- Added package structure verification tests
- Fixed MCP server import tests
- Added pytest configuration in pyproject.toml

Documentation

Updated README.md - PyPI badges, reordered installation options
ROADMAP.md - Comprehensive roadmap with task-based approach
Installation guides - Simplified with PyPI as primary method
Testing documentation - How to run full test suite

Changed

Package Structure

Moved to src/ layout:
- src/skill_seekers/ - Main package
- src/skill_seekers/cli/ - CLI tools
- src/skill_seekers/mcp/ - MCP server
Import paths updated - All imports use proper package structure
Entry points configured - All CLI tools available as commands

Import Fixes

Fixed merge_sources.py - Corrected conflict_detector import (.conflict_detector)
Fixed MCP server tests - Updated to use skill_seekers.mcp.server imports
Fixed test paths - All tests updated for src/ layout

Fixed

Critical Bugs

Import path errors - Fixed relative imports in CLI modules
MCP test isolation - Added proper MCP availability checks
Package installation - Resolved entry point conflicts
Dependency resolution - All dependencies properly specified

Test Improvements

17 test fixes - Updated for modern package structure
MCP test guards - Proper skipif decorators for MCP tests
CLI test updates - Accept both exit codes 0 and 2 for help
Path validation - Tests verify correct package structure

Technical Details

Build System

Build backend: setuptools.build_meta
Build command: uv build
Publish command: uv publish
Distribution formats: wheel + source tarball

Dependencies

Core: requests, beautifulsoup4, PyGithub, mcp, httpx
PDF: PyMuPDF, Pillow, pytesseract
Dev: pytest, pytest-cov, pytest-anyio, mypy
MCP: mcp package for Claude Code integration

Migration Guide

For Users

Old way:

git clone https://github.com/yusufkaraaslan/Skill_Seekers.git
cd Skill_Seekers
pip install -r requirements.txt
python3 cli/doc_scraper.py --config configs/react.json

New way:

pip install skill-seekers
skill-seekers scrape --config configs/react.json

For Developers

Update imports: from cli.* → from skill_seekers.cli.*
Use pip install -e ".[dev]" for development
Run tests: python -m pytest
Entry points instead of direct script execution

Breaking Changes

CLI interface changed - Use skill-seekers command instead of python3 cli/...
Import paths changed - Package now at skill_seekers.* instead of cli.*
Installation method changed - PyPI recommended over git clone

Deprecations

Direct script execution - Still works but deprecated (use skill-seekers command)
Old import patterns - Legacy imports still work but will be removed in v3.0

Compatibility

Python 3.10+ required
Backward compatible - Old scripts still work with legacy CLI
Config files - No changes required
Output format - No changes to generated skills

[1.3.0] - 2025-10-26

Added - Refactoring & Performance Improvements

Async/Await Support for Parallel Scraping (2-3x performance boost)
- --async flag to enable async mode
- async def scrape_page_async() method using httpx.AsyncClient
- async def scrape_all_async() method with asyncio.gather()
- Connection pooling for better performance
- asyncio.Semaphore for concurrency control
- Comprehensive async testing (11 new tests)
- Full documentation in ASYNC_SUPPORT.md
- Performance: ~55 pages/sec vs ~18 pages/sec (sync)
- Memory: 40 MB vs 120 MB (66% reduction)
Python Package Structure (Phase 0 Complete)
- cli/__init__.py - CLI tools package with clean imports
- skill_seeker_mcp/__init__.py - MCP server package (renamed from mcp/)
- skill_seeker_mcp/tools/__init__.py - MCP tools subpackage
- Proper package imports: from cli import constants
Centralized Configuration Module
- cli/constants.py with 18 configuration constants
- DEFAULT_ASYNC_MODE, DEFAULT_RATE_LIMIT, DEFAULT_MAX_PAGES
- Enhancement limits, categorization scores, file limits
- All magic numbers now centralized and configurable
Code Quality Improvements
- Converted 71 print() statements to proper logging calls
- Added type hints to all DocToSkillConverter methods
- Fixed all mypy type checking issues
- Installed types-requests for better type safety
Multi-variant llms.txt detection: downloads all 3 variants (full, standard, small)
Automatic .txt → .md file extension conversion
No content truncation: preserves complete documentation
detect_all() method for finding all llms.txt variants
get_proper_filename() for correct .md naming

Changed

_try_llms_txt() now downloads all available variants instead of just one
Reference files now contain complete content (no 2500 char limit)
Code samples now include full code (no 600 char limit)
Test count increased from 207 to 299 (92 new tests)
All print() statements replaced with logging (logger.info, logger.warning, logger.error)
Better IDE support with proper package structure
Code quality improved from 5.5/10 to 6.5/10

Fixed

File extension bug: llms.txt files now saved as .md
Content loss: 0% truncation (was 36%)
Test isolation issues in test_async_scraping.py (proper cleanup with try/finally)
Import issues: no more sys.path.insert() hacks needed
.gitignore: added test artifacts (.pytest_cache, .coverage, htmlcov, etc.)

1.2.0 - 2025-10-23

🚀 PDF Advanced Features Release

Major enhancement to PDF extraction capabilities with Priority 2 & 3 features.

Added

Priority 2: Support More PDF Types

OCR Support for Scanned PDFs
- Automatic text extraction from scanned documents using Tesseract OCR
- Fallback mechanism when page text < 50 characters
- Integration with pytesseract and Pillow
- Command: --ocr flag
- New dependencies: Pillow==11.0.0, pytesseract==0.3.13
Password-Protected PDF Support
- Handle encrypted PDFs with password authentication
- Clear error messages for missing/wrong passwords
- Secure password handling
- Command: --password PASSWORD flag
Complex Table Extraction
- Extract tables from PDFs using PyMuPDF's table detection
- Capture table data as 2D arrays with metadata (bbox, row/col count)
- Integration with skill references in markdown format
- Command: --extract-tables flag

Priority 3: Performance Optimizations

Parallel Page Processing
- 3x faster PDF extraction using ThreadPoolExecutor
- Auto-detect CPU count or custom worker specification
- Only activates for PDFs with > 5 pages
- Commands: --parallel and --workers N flags
- Benchmarks: 500-page PDF reduced from 4m 10s to 1m 15s
Intelligent Caching
- In-memory cache for expensive operations (text extraction, code detection, quality scoring)
- 50% faster on re-runs
- Command: --no-cache to disable (enabled by default)

New Documentation

docs/PDF_ADVANCED_FEATURES.md (580 lines)
- Complete usage guide for all advanced features
- Installation instructions
- Performance benchmarks showing 3x speedup
- Best practices and troubleshooting
- API reference with all parameters

Testing

New test file: tests/test_pdf_advanced_features.py (568 lines, 26 tests)
- TestOCRSupport (5 tests)
- TestPasswordProtection (4 tests)
- TestTableExtraction (5 tests)
- TestCaching (5 tests)
- TestParallelProcessing (4 tests)
- TestIntegration (3 tests)
Updated: tests/test_pdf_extractor.py (23 tests fixed and passing)
Total PDF tests: 49/49 PASSING ✅ (100% pass rate)

Changed

Enhanced cli/pdf_extractor_poc.py with all advanced features
Updated requirements.txt with new dependencies
Updated README.md with PDF advanced features usage
Updated docs/TESTING.md with new test counts (142 total tests)

Performance Improvements

3.3x faster with parallel processing (8 workers)
1.7x faster on re-runs with caching enabled
Support for unlimited page PDFs (no more 500-page limit)

Dependencies

Added Pillow==11.0.0 for image processing
Added pytesseract==0.3.13 for OCR support
Tesseract OCR engine (system package, optional)

1.1.0 - 2025-10-22

🌐 Documentation Scraping Enhancements

Major improvements to documentation scraping with unlimited pages, parallel processing, and new configs.

Added

Unlimited Scraping & Performance

Unlimited Page Scraping - Removed 500-page limit, now supports unlimited pages
Parallel Scraping Mode - Process multiple pages simultaneously for faster scraping
Dynamic Rate Limiting - Smart rate limit control to avoid server blocks
CLI Utilities - New helper scripts for common tasks

New Configurations

Ansible Core 2.19 - Complete Ansible documentation config
Claude Code - Documentation for this very tool!
Laravel 9.x - PHP framework documentation

Testing & Quality

Comprehensive test coverage for CLI utilities
Parallel scraping test suite
Virtual environment setup documentation
Thread-safety improvements

Fixed

Thread-safety issues in parallel scraping
CLI path references across all documentation
Flaky upload_skill tests
MCP server streaming subprocess implementation

Changed

All CLI examples now use cli/ directory prefix
Updated documentation structure
Enhanced error handling

1.0.0 - 2025-10-19

🎉 First Production Release

This is the first production-ready release of Skill Seekers with complete feature set, full test coverage, and comprehensive documentation.

Added

Smart Auto-Upload Feature

New upload_skill.py CLI tool for automatic API-based upload
Enhanced package_skill.py with --upload flag
Smart API key detection with graceful fallback
Cross-platform folder opening in utils.py
Helpful error messages instead of confusing errors

MCP Integration Enhancements

9 MCP tools (added upload_skill tool)
mcp__skill-seeker__upload_skill - Upload .zip files to Claude automatically
Enhanced package_skill tool with smart auto-upload parameter
Updated all MCP documentation to reflect 9 tools

Documentation Improvements

Updated README with version badge (v1.0.0)
Enhanced upload guide with 3 upload methods
Updated MCP setup guide with all 9 tools
Comprehensive test documentation (14/14 tests)
All references to tool counts corrected

Fixed

Missing import os in mcp/server.py
package_skill.py exit code behavior (now exits 0 when API key missing)
Improved UX with helpful messages instead of errors

Changed

Test count badge updated (96 → 14 passing)
All documentation references updated to 9 tools

Testing

CLI Tests: 8/8 PASSED ✅
MCP Tests: 6/6 PASSED ✅
Total: 14/14 PASSED (100%)

0.4.0 - 2025-10-18

Added

Large Documentation Support (40K+ Pages)

Config splitting functionality for massive documentation sites
Router/hub skill generation for intelligent query routing
Checkpoint/resume feature for long scrapes
Parallel scraping support for faster processing
4 split strategies: auto, category, router, size

New CLI Tools

split_config.py - Split large configs into focused sub-skills
generate_router.py - Generate router/hub skills
package_multi.py - Package multiple skills at once

New MCP Tools

split_config - Split large documentation via MCP
generate_router - Generate router skills via MCP

Documentation

New docs/LARGE_DOCUMENTATION.md guide
Example config: godot-large-example.json (40K pages)

Changed

MCP tool count: 6 → 8 tools
Updated documentation for large docs workflow

0.3.0 - 2025-10-15

Added

MCP Server Integration

Complete MCP server implementation (mcp/server.py)
6 MCP tools for Claude Code integration:
- list_configs
- generate_config
- validate_config
- estimate_pages
- scrape_docs
- package_skill

Setup & Configuration

Automated setup script (setup_mcp.sh)
MCP configuration examples
Comprehensive MCP setup guide (docs/MCP_SETUP.md)
MCP testing guide (docs/TEST_MCP_IN_CLAUDE_CODE.md)

Testing

31 comprehensive unit tests for MCP server
Integration tests via Claude Code MCP protocol
100% test pass rate

Documentation

Complete MCP integration documentation
Natural language usage examples
Troubleshooting guides

Changed

Restructured project as monorepo with CLI and MCP server
Moved CLI tools to cli/ directory
Added MCP server to mcp/ directory

0.2.0 - 2025-10-10

Added

Testing & Quality

Comprehensive test suite with 71 tests
100% test pass rate
Test coverage for all major features
Config validation tests

Optimization

Page count estimator (estimate_pages.py)
Framework config optimizations with start_urls
Better URL pattern coverage
Improved scraping efficiency

New Configs

Kubernetes documentation config
Tailwind CSS config
Astro framework config

Changed

Optimized all framework configs
Improved categorization accuracy
Enhanced error messages

0.1.0 - 2025-10-05

Added

Initial Release

Basic documentation scraper functionality
Manual skill creation
Framework configs (Godot, React, Vue, Django, FastAPI)
Smart categorization system
Code language detection
Pattern extraction
Local and API-based enhancement options
Basic packaging functionality

Core Features

BFS traversal for documentation scraping
CSS selector-based content extraction
Smart categorization with scoring
Code block detection and formatting
Caching system for scraped data
Interactive mode for config creation

Documentation

README with quick start guide
Basic usage documentation
Configuration file examples

Release Links

v1.2.0 - PDF Advanced Features
v1.1.0 - Documentation Scraping Enhancements
v1.0.0 - Production Release
v0.4.0 - Large Documentation Support
v0.3.0 - MCP Integration

Version History Summary

Version	Date	Highlights
1.2.0	2025-10-23	📄 PDF advanced features: OCR, passwords, tables, 3x faster
1.1.0	2025-10-22	🌐 Unlimited scraping, parallel mode, new configs (Ansible, Laravel)
1.0.0	2025-10-19	🚀 Production release, auto-upload, 9 MCP tools
0.4.0	2025-10-18	📚 Large docs support (40K+ pages)
0.3.0	2025-10-15	🔌 MCP integration with Claude Code
0.2.0	2025-10-10	🧪 Testing & optimization
0.1.0	2025-10-05	🎬 Initial release

164 KiB Raw Blame History Unescape Escape

Changelog

Unreleased

Added

Fixed

[3.4.0] - 2026-03-21

Added

Fixed

Changed

[3.3.0] - 2026-03-16

Supported Source Types (17)

Added

10 New Skill Source Types (17 total)

EPUB Unified Pipeline Integration

Unified Skill Builder — Generic Merge System

Config Validator Expansion

Source Detection Improvements

MCP Server Integration

CLI Wiring

Sync Config Command

Workflow & Documentation

Performance

Fixed

[3.2.0] - 2026-03-01

🎬 Video Tutorial Scraping Pipeline (BETA)

Added

Video Pipeline Core (skill-seekers video)

Video --setup: GPU Auto-Detection & Dependency Installation

Microsoft Word (.docx) Support

Other Additions

Fixed

Video Pipeline OCR Quality Fixes (6)

Video Pipeline Fixes (15)

Scraper & Quality Fixes

Changed

[3.1.3] - 2026-02-24

🐛 Hotfix — Explicit Chunk Flags & Argument Pipeline Cleanup

Fixed

Changed

[3.1.2] - 2026-02-24

🔧 Fix create Command Argument Forwarding, Gemini Model, and Enhance Dispatcher

Fixed

Added

Changed

[3.1.1] - 2026-02-23

🐛 Hotfix

Fixed

Changed

[3.1.0] - 2026-02-23

🎯 "Unified CLI & Developer Experience" — Feature Release

Added

Unified create Command

Enhancement Workflow Preset System

Multiple --enhance-workflow Flags from CLI

Smart Enhancement Dispatcher (skill-seekers enhance)

Unified Document Parser System

Local Source Support in Unified Scraper

CLI Flag Parity Across All Commands

Fixed

Changed

Tests

Config Repository (skill-seekers-configs)

[3.0.0] - 2026-02-10

🚀 "Universal Intelligence Platform" - Major Release

Highlights

Added - Platform Adaptors (16 Total)

RAG & Vector Databases (8)

AI Platforms (3)

AI Coding Assistants (4)

Generic (1)

Added - MCP Tools (26 Total)

Config Tools (3)

Scraping Tools (8)

Packaging Tools (4)

Source Tools (5)

Splitting Tools (2)

Vector DB Tools (4)

Added - Cloud Storage

Added - CI/CD Support

GitHub Action

164 KiB

Raw Blame History

Video Pipeline Core (`skill-seekers video`)

Video `--setup`: GPU Auto-Detection & Dependency Installation

🔧 Fix `create` Command Argument Forwarding, Gemini Model, and Enhance Dispatcher

Unified `create` Command

Multiple `--enhance-workflow` Flags from CLI

Smart Enhancement Dispatcher (`skill-seekers enhance`)