Skip to main content
Glama
srwlli

Documentation Generator MCP Server

by srwlli
handoff-workflow.txt5.77 kB
================================================================================ IDEA LOGGING: /HANDOFF WORKFLOW ================================================================================ IDEA: SLASH COMMAND FOR AGENT HANDOFF AUTOMATION DATE: 2025-10-16 [Current Session] STATUS: Proposed DESCRIPTION: Create a /handoff slash command that automatically generates agent context files (claude.md) at project and phase levels to enable seamless continuity when agents change mid-project. Streamlines the handoff process documented in agent-context-workflow.txt. CONTEXT: Mid-project agent changes are common in development workflows. Currently, creating context files for handoffs is a manual process requiring developers to copy templates and fill them out. This creates friction and inconsistency. A /handoff command would automate context capture, reduce handoff time from 20-30 minutes to 2-3 minutes, and ensure consistency across all handoffs. RELATIONSHIP TO MCP: Core docs-mcp workflow tool that complements existing planning and documentation tools. Integrates with /create-plan, /analyze-for-planning, and /generate-docs to provide mid-project continuity. Extends the MCP ecosystem with agent-to-agent handoff capabilities. RELATED WORK: - coderef/future/agent-context-workflow.txt (defines manual handoff process) - coderef/future/idea-logging-workflow.txt (this template source) - /create-plan (creates plan.json that handoff references) - /analyze-for-planning (provides project analysis for context) - coderef/working/{feature-name}/ structure (target for context files) PRELIMINARY THOUGHTS: - Command should create both project-level and phase-level claude.md files - Auto-detect current phase from plan.json or working directory structure - Extract project goals, architecture, and decisions from plan.json - Parse git history for recent changes and current progress - Include option to create minimal vs. full context files - Should intelligently populate "What's Been Completed" from git commits - Could integrate with TODO tracking for "What's Currently In Progress" - Generate file references (hot spots) from git diff statistics - Support both initial handoff (project start) and mid-phase handoff - Consider auto-populating "Notes from Last Agent" from conversation context IMPLEMENTATION CONSIDERATIONS: 1. MCP Tool: `generate_handoff_context` - Input: project_path, feature_name, phase_number (optional), mode (full|minimal) - Output: Created files (project-level and/or phase-level claude.md) 2. Data Sources for Auto-Population: - plan.json → Project goals, architecture, phase objectives - analysis.json → Tech stack, dependencies, file structure - Git history → Recent changes, progress tracking - Current directory structure → Phase detection - Existing claude.md files → Update vs. create logic 3. Template Selection: - Full mode: Use complete templates from agent-context-workflow.txt - Minimal mode: Use minimal template for quick capture - Auto-detect: Choose based on project complexity 4. Smart Features: - Parse plan.json tasks and map to completed/in-progress sections - Extract blocker information from TODO comments in code - Generate "How to Get Started" from README and package.json - Auto-link to related files (plan.json, analysis.json, etc.) - Timestamp all updates automatically - Preserve "Notes from Last Agent" history across handoffs NEXT STEPS: - [ ] Use /gather-context to capture detailed requirements * Interview questions: What handoff scenarios to support? * User stories: Who uses /handoff and when? * Acceptance criteria: What makes a successful handoff? - [ ] Move to planning phase when ready - [ ] Create implementation plan with /create-plan - [ ] Archive to /archived/ once executed POTENTIAL CHALLENGES: - Phase detection: How to reliably detect current phase without explicit input? - Stale data: How to handle outdated plan.json or analysis.json? - Merge conflicts: How to update existing claude.md without overwriting manual edits? - Context accuracy: How to validate auto-generated context is correct? USER EXPERIENCE FLOW: ``` # Scenario 1: Initial project handoff (agent starting work) /handoff # → Creates project-level claude.md # → Creates phase-1/claude.md if plan.json exists # → Populates from plan.json and analysis.json # → Returns: "Handoff context created at coderef/working/{feature}/claude.md" # Scenario 2: Mid-phase handoff (agent switching mid-work) /handoff --phase 2 --mode minimal # → Updates phase-2/claude.md with current state # → Extracts progress from git commits # → Adds "Notes from Last Agent" section # → Returns: "Phase 2 handoff context updated" # Scenario 3: Phase transition handoff (completing phase, starting next) /handoff --transition # → Marks current phase claude.md as "Complete" # → Creates next phase claude.md # → Updates project-level claude.md with current phase # → Returns: "Phase 2 → 3 transition handoff complete" ``` BENEFITS: - Fast context capture: 2-3 minutes vs. 20-30 minutes manual - Consistency: All handoffs follow same template structure - Reduced errors: Auto-population prevents missing critical info - Better continuity: New agents can start immediately with clear context - Historical tracking: Preserves decision history across agents - Low friction: Single command vs. copying templates and filling manually SUCCESS METRICS: - Agent handoff time reduced by 80% (20-30 min → 2-3 min) - 100% handoff consistency (all use same template structure) - Zero missing critical context fields - New agent productive within 5-10 minutes (vs. 1+ hour) ================================================================================

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/srwlli/docs-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server