Documentation Generator MCP Server

================================================================================ AGENT CONTEXT WORKFLOW - MID-PROJECT CONTINUITY ================================================================================ PURPOSE: Provide rapid context for new agents joining mid-project. Enable seamless handoff and continuity by documenting current state, decisions, blockers, and next immediate actions. ================================================================================ FILE PLACEMENT STRATEGY ================================================================================ 1. PROJECT ROOT LEVEL: Location: /{working|archived}/{feature-name}/claude.md Scope: High-level project overview and architecture Updates: At project start and major milestones 2. PHASE-LEVEL FILES: Location: /{working|archived}/{feature-name}/phase-{N}/claude.md Scope: Current phase status, blockers, immediate next steps Updates: Before phase transition or when agent changes 3. FOLDER STRUCTURE: /working/{feature-name}/ ├── claude.md (PROJECT OVERVIEW) ├── plan.json ├── analysis.json ├── phase-1/ │ ├── claude.md (PHASE 1 CONTEXT) │ └── [implementation files] ├── phase-2/ │ ├── claude.md (PHASE 2 CONTEXT) │ └── [implementation files] └── phase-3/ ├── claude.md (PHASE 3 CONTEXT) └── [implementation files] ================================================================================ PROJECT-LEVEL CLAUDE.MD TEMPLATE ================================================================================ # Project: {Feature Name} **Status:** [Planning/In Progress/Testing/Ready for Merge/Completed] **Last Updated:** YYYY-MM-DD HH:MM **Current Agent:** [Name/ID if applicable] ## Quick Summary [1-2 sentence overview of what this feature does] ## Project Goals - [Goal 1] - [Goal 2] - [Goal 3] ## Architecture Overview [Paragraph describing overall system design, key components, tech stack] ## Key Decisions Made - **Decision 1:** [What] → [Why] → [Impact] - **Decision 2:** [What] → [Why] → [Impact] - **Decision 3:** [What] → [Why] → [Impact] ## Current Phase **Phase:** [N of total] - [Phase Name] **Progress:** [% complete or status description] **Started:** YYYY-MM-DD **Est. Completion:** YYYY-MM-DD ## Known Blockers - [ ] Blocker 1: [Description] - Assigned to [person/status] - [ ] Blocker 2: [Description] - Assigned to [person/status] ## Dependencies - [External service/library]: [version/status] - [Database/config]: [current state] - [Related features]: [link to other projects] ## File Structure - `/src/` - Main implementation - `/tests/` - Test files - `/docs/` - Additional documentation - `/phase-N/` - Phase-specific work ## How to Get Started (New Agent) 1. Read plan.json for full implementation details 2. Review this file and current phase context 3. Check latest phase/claude.md for immediate next steps 4. Review related decision logs in phase directories 5. [Any setup instructions: npm install, build, etc.] ## References - Plan: plan.json - Analysis: analysis.json - Related: [link to idea log or related projects] ================================================================================ PHASE-LEVEL CLAUDE.MD TEMPLATE ================================================================================ # Phase {N}: {Phase Name} **Status:** [Not Started/In Progress/Blocked/Complete] **Assigned To:** [Agent/Person] **Started:** YYYY-MM-DD **Completed:** [YYYY-MM-DD or N/A] ## Phase Objective [Clear, specific goal for this phase] ## What's Been Completed ✓ - [x] Task 1: [Description] - [x] Task 2: [Description] - [x] Task 3: [Description] ## What's Currently In Progress 🔄 - [ ] Task A: [Description] - [% complete] - [ ] Task B: [Description] - [% complete] ## What's Next (Priority Order) 1. [ ] Next Task 1: [Description] - [Est. time] 2. [ ] Next Task 2: [Description] - [Est. time] 3. [ ] Next Task 3: [Description] - [Est. time] ## Critical Context for Continuity ### Recent Decisions - **Decision A:** [What changed and why] - **Decision B:** [What changed and why] ### Active Blockers - **Blocker:** [Issue] - [Attempted solutions] - [Owner/Status] - **Blocker:** [Issue] - [Attempted solutions] - [Owner/Status] ### Code Hot Spots - File: `src/component.ts:45` - [Why this is important] - File: `src/service.ts:120` - [Why this is important] ### Test Status - Unit tests: [Passing/[X] failing - details] - Integration tests: [Passing/[X] failing - details] - E2E tests: [Status or N/A] ## Environment Setup ``` [Any commands or setup specific to this phase] npm install npm run build npm run dev ``` ## How to Continue From Here 1. [First immediate action] 2. [Second immediate action] 3. [Third immediate action] 4. Refer to Phase {N+1} plan when ready to transition ## Notes from Last Agent [Any lessons learned, gotchas, or tips for the next agent] ## Quick Links - Previous phase: ../phase-{N-1}/claude.md - Next phase plan: ../phase-{N+1}/claude.md (if exists) - Project overview: ../claude.md ================================================================================ MINIMAL CONTEXT FILE (For Quick Capture) ================================================================================ When time is short, a minimal version can be used: --- # {Feature} - Phase {N} **Status:** [Status] **Last Update:** YYYY-MM-DD ## Where We Are [2-3 sentences about current state] ## What's Blocked - [Blocker 1] - [Blocker 2] ## Do Next 1. [Next action 1] 2. [Next action 2] ## Hot Files - path/to/file1.ts - path/to/file2.ts --- ================================================================================ WORKFLOW: CREATING AND UPDATING CONTEXT FILES ================================================================================ 1. PROJECT INITIATION: - Create project-level claude.md at /working/{feature-name}/claude.md - Fill with plan.json high-level summary - Set initial status to "Planning" 2. PHASE START: - Create phase-level claude.md at /working/{feature-name}/phase-{N}/ - Copy objectives from plan.json phase section - Set status to "In Progress" 3. DURING PHASE: - Update phase claude.md daily/weekly with progress - Mark completed tasks with ✓ - Document blockers as they arise - Note any decision changes 4. PHASE TRANSITION: - Update phase claude.md with "Complete" status - Create next phase claude.md with initial content - Update project-level claude.md with current phase 5. PROJECT COMPLETION: - Mark all files with "Complete" status - Archive project to /archived/{feature-name}/ - Preserve all claude.md files for historical reference 6. AGENT HANDOFF: - Before passing to new agent: * Update current phase claude.md with latest status * Document any new blockers or decisions * Add "Notes from Last Agent" section * Include specific file/line references if applicable ================================================================================ QUICK REFERENCE: WHAT GOES WHERE ================================================================================ PROJECT-LEVEL (claude.md): ✓ Feature overview and goals ✓ Architecture and tech stack ✓ Major decisions made ✓ Current phase and overall progress ✓ Project-level blockers and dependencies ✗ Day-to-day task status (→ Phase level) ✗ Specific line numbers or file edits (→ Phase level) PHASE-LEVEL (claude.md): ✓ Phase-specific objective ✓ Completed tasks with details ✓ Current in-progress work ✓ Next immediate actions ✓ Phase-level blockers ✓ Code hot spots and gotchas ✓ Test status ✓ Environment setup commands ✗ Project-wide decisions (→ Project level) ✗ Long-term roadmap (→ Project level) ================================================================================ BENEFITS ================================================================================ For New Agents: - Rapid context acquisition (5-10 min vs. 1+ hour) - Clear next steps without searching - Understanding of past decisions and why they were made - Awareness of current blockers and gotchas - Specific code locations to focus on For Projects: - Continuity preserved across agent changes - Reduced ramp-up time - Clear progress tracking - Historical record of decision-making - Faster debugging (code hot spots documented) For Teams: - Better handoff between developers/agents - Clearer communication of project state - Easier code review with context - Reduced duplicate work and mistakes ================================================================================