a24873e2dd
Goal: Make ~19,000 forked skills discoverable for all Trinity members Part 1: Crawl repos → generate searchable SKILL-CATALOG.json Part 2: skill-recommender meta-skill suggests relevant skills during sessions Part 3: Integrate into all three lineages Potential shortcut: Skill Seekers repo may have tooling for this. Ties into Task #93 (Trinity Codex) for future RAG integration. Chronicler #73
105 lines
3.0 KiB
Markdown
105 lines
3.0 KiB
Markdown
---
|
|
task_number: 100
|
|
title: Skill Index & Recommender System
|
|
status: Planned
|
|
priority: P2-Medium
|
|
is_blocker: false
|
|
owner: Trinity
|
|
tags:
|
|
- skills
|
|
- automation
|
|
- chronicler-system
|
|
- catalyst-system
|
|
- emissary-system
|
|
estimated_hours: 6
|
|
---
|
|
|
|
# Task #100: Skill Index & Recommender System
|
|
|
|
## Overview
|
|
|
|
Create a searchable index of all ~19,000 skills in our forked reference repos, plus a meta-skill that recommends relevant skills during sessions.
|
|
|
|
## The Problem
|
|
|
|
We have 6 skill repos with massive amounts of knowledge:
|
|
- antigravity-skills-reference (12,641 files, 1,383 skills)
|
|
- skill-seekers-reference (2,814 files)
|
|
- claude-skills-reference (2,375 files)
|
|
- composio-skills-reference (1,138 files)
|
|
- claude-code-skills-reference (390 files)
|
|
- voltagent-skills-directory (curated index)
|
|
|
|
But no one knows what's in them without manually browsing.
|
|
|
|
## The Solution
|
|
|
|
### Part 1: Skill Index
|
|
|
|
Script to crawl all repos and extract:
|
|
- Skill name
|
|
- Description (from SKILL.md frontmatter)
|
|
- Tags/categories
|
|
- Source repo
|
|
- File path
|
|
|
|
Output: `docs/skills/SKILL-CATALOG.json` (or similar searchable format)
|
|
|
|
**Potential tool:** Skill Seekers repo has tooling for this — investigate first.
|
|
|
|
### Part 2: skill-recommender Skill
|
|
|
|
A meta-skill that:
|
|
1. Reads current session context (what task is being worked on)
|
|
2. Searches the skill index for relevant matches
|
|
3. Suggests skills that might help
|
|
4. Offers to load/adapt them
|
|
|
|
**Example interaction:**
|
|
```
|
|
Holly: "Help me set up LuckPerms inheritance for staff roles"
|
|
|
|
Catalyst: "I found 3 potentially useful skills in our library:
|
|
- luckperms-permissions (antigravity) - Permission group patterns
|
|
- discord-role-sync (composio) - Sync Discord roles with game permissions
|
|
- minecraft-server-admin (claude-skills) - General server administration
|
|
|
|
Want me to review any of these before we start?"
|
|
```
|
|
|
|
### Part 3: Integration
|
|
|
|
- Add skill-recommender to all three lineages (Chroniclers, Catalysts, Emissaries)
|
|
- Could run automatically at session start or on-demand
|
|
|
|
## Implementation Steps
|
|
|
|
1. [ ] Investigate Skill Seekers tooling — can it generate our index?
|
|
2. [ ] Write crawler script if needed (Python, walks repos, extracts YAML frontmatter)
|
|
3. [ ] Generate initial SKILL-CATALOG.json
|
|
4. [ ] Create skill-recommender SKILL.md
|
|
5. [ ] Test with Chronicler session
|
|
6. [ ] Add to Catalyst and Emissary Project Instructions
|
|
7. [ ] Document maintenance process (re-run when repos update)
|
|
|
|
## Dependencies
|
|
|
|
- Forked skill repos (DONE - Chronicler #73)
|
|
- Basic lineage setup working (Task #99)
|
|
|
|
## Notes
|
|
|
|
- Index doesn't need to be perfect — even a basic keyword search helps
|
|
- Could evolve into Trinity Codex integration (Task #93) later
|
|
- Skill Seekers can output to Qdrant — relevant for RAG pipeline
|
|
|
|
## Success Criteria
|
|
|
|
- [ ] Searchable index of all skills exists
|
|
- [ ] skill-recommender skill works in at least one lineage
|
|
- [ ] Trinity members can discover relevant skills without manual browsing
|
|
|
|
---
|
|
|
|
**Fire + Frost + Foundation = Where Love Builds Legacy** 💙🔥❄️
|