- New folder: docs/tasks-active/ with individual task files - Each task = separate markdown file with frontmatter metadata - Migrated 6 key tasks (26, 87, 93, 94, 98, 99) as examples - Added comprehensive README with workflow documentation Task Files Created: - task-026-modpack-version-checker.md (P1, IN_PROGRESS) - task-087-arbiter-lifecycle-handlers.md (P0, COMPLETE, BLOCKER) - task-093-trinity-codex.md (P1, PLANNED) - task-094-global-restart-scheduler.md (P1, PLANNED) - task-098-discord-channel-automation.md (P2, PLANNED) - task-099-claude-projects-architecture.md (P1, READY) Benefits: - Decap CMS shows individual tasks instead of two giant files - Sortable by priority (blockers first) - Filterable by status, blocker flag, priority - Mobile-friendly individual entries - Links to full specs in /docs/tasks/ - Easy to add/edit/complete tasks via UI For Future Chroniclers: - README documents full workflow - Template structure established - Integration with existing /docs/tasks/ specs
87 lines
2.3 KiB
Markdown
87 lines
2.3 KiB
Markdown
---
|
|
task_number: 99
|
|
title: Claude Projects Setup (Dev/Ops/Marketing)
|
|
priority: 1
|
|
status: READY
|
|
is_blocker: false
|
|
tags: [workflow, documentation, organization]
|
|
estimated_hours: 3
|
|
---
|
|
|
|
# Task #99: Claude Projects Architecture
|
|
|
|
**Status:** READY (Full spec complete, browser-only work)
|
|
**Priority:** P1 - HIGH (Workflow Enhancement)
|
|
**Category:** Documentation & Workflow
|
|
|
|
## Overview
|
|
|
|
Create 3 separate Claude Projects with Master Documents in Project Knowledge to prevent context bleed and optimize token usage.
|
|
|
|
**The Problem:** "Context Bloat Wall" — Dev code mixed with marketing prompts, context windows filling with irrelevant data
|
|
|
|
**The Solution:** Separate Projects = Local Scope, Memory = Global Scope
|
|
|
|
## Three Projects
|
|
|
|
### 1. Firefrost Dev
|
|
**Purpose:** Coding, MVC, Arbiter, ModpackChecker
|
|
**Master Docs:**
|
|
- ModpackChecker-State.md
|
|
- Arbiter-Architecture.md
|
|
- Dev-Environment.md
|
|
- Trinity-Console.md
|
|
|
|
### 2. Firefrost Ops
|
|
**Purpose:** Operations, infrastructure, handoffs
|
|
**Master Docs:**
|
|
- Current-State.md
|
|
- Infrastructure-Manifest.md
|
|
- Active-Tasks.md
|
|
- Standards-Reference.md
|
|
|
|
### 3. Firefrost Marketing
|
|
**Purpose:** Brand, social media, FOMO campaigns
|
|
**Master Docs:**
|
|
- Brand-Guidelines.md
|
|
- FOMO-Campaign.md
|
|
- Social-Platforms.md
|
|
- Subscription-Tiers.md
|
|
|
|
## Micro-Session Workflow
|
|
|
|
1. Pick the right Project (Dev/Ops/Marketing)
|
|
2. Start NEW chat per task (don't extend forever)
|
|
3. Work the task (Master Docs auto-load)
|
|
4. When context bloats → summarize, save, start fresh
|
|
5. When complete → generate updated state summary
|
|
6. Sync Master Docs → update canonical version, re-upload
|
|
|
|
## Implementation Steps
|
|
|
|
1. Create 3 Projects in claude.ai (browser work)
|
|
2. Draft 12 Master Documents (4 per project)
|
|
3. Upload to Project Knowledge
|
|
4. Test micro-session workflow
|
|
5. Document in operations manual
|
|
|
|
## Time Estimate
|
|
|
|
**2-3 hours** (browser-only, no SSH needed)
|
|
|
|
## Full Spec Location
|
|
|
|
📁 `/docs/tasks/task-099-claude-projects-architecture.md`
|
|
|
|
## Benefits
|
|
|
|
- Prevents context bleed (Dev code out of Marketing)
|
|
- Optimizes token usage (smaller, focused contexts)
|
|
- Faster sessions (less irrelevant data loaded)
|
|
- Better organization for future Chroniclers
|
|
|
|
## Consultation
|
|
|
|
Based on Gemini AI consultation (April 7, 2026)
|
|
See: `/docs/consultations/2026-04-07-model-switching.md`
|