AgentSkillsCN

ai-memory

ai-memory

SKILL.md

AI Memory System Skill

Purpose: Manage persistent memory files (STATUS.md, ROADMAP.md, DECISIONS.md, JOURNAL.md) for AI-assisted development across sessions.

Use this skill when:

  • Starting a new AI Friday session
  • Ending a session and need to capture progress
  • Making significant architectural decisions
  • Discovering important learnings
  • Initializing memory system for a new project or team

Available Commands

/memory-init - Initialize Memory System

Creates the core memory files from templates.

When to use: First time setting up the memory system for a project or team branch.

What it does:

  1. Creates STATUS.md (current state tracking)
  2. Creates ROADMAP.md (milestone checklists)
  3. Creates DECISIONS.md (decision log)
  4. Creates JOURNAL.md (learning documentation)
  5. Ensures CLAUDE.md exists (or creates from template)

Usage:

code
User: /memory-init

/memory-start - Start Session

Reads all memory files and provides situational awareness.

When to use: At the beginning of every AI Friday session.

What it does:

  1. Reads CLAUDE.md (project context)
  2. Reads STATUS.md (current state)
  3. Reads ROADMAP.md (what's next)
  4. Reads DECISIONS.md (decision history)
  5. Summarizes current state and next steps

Usage:

code
User: /memory-start

/memory-update - Update Status

Updates STATUS.md with current progress and next steps.

When to use: At the end of a session, or when switching tasks.

What it does:

  1. Reads current STATUS.md
  2. Asks what was completed
  3. Asks what's in progress
  4. Asks about blockers
  5. Asks about next steps
  6. Updates STATUS.md with new information

Usage:

code
User: /memory-update

/memory-decision - Log Decision

Adds a decision entry to DECISIONS.md.

When to use: After making a significant architectural, technical, or process decision.

What it does:

  1. Asks for decision details
  2. Prompts for alternatives considered
  3. Prompts for rationale
  4. Prompts for tradeoffs
  5. Formats and appends to DECISIONS.md

Usage:

code
User: /memory-decision

/memory-learn - Capture Learning

Adds a learning entry to JOURNAL.md.

When to use: When you discover something non-obvious or worth remembering.

What it does:

  1. Asks what was discovered
  2. Asks for context/why it matters
  3. Asks for solution/pattern
  4. Requests code examples (if applicable)
  5. Formats and appends to JOURNAL.md

Usage:

code
User: /memory-learn

/memory-roadmap - Update Roadmap

Marks tasks complete or adds new tasks to ROADMAP.md.

When to use: When completing milestones or tasks.

What it does:

  1. Shows current ROADMAP.md
  2. Asks which tasks to mark complete
  3. Asks if new tasks should be added
  4. Updates ROADMAP.md accordingly

Usage:

code
User: /memory-roadmap

/memory-summary - Session Summary

Generates a summary of the session for STATUS.md.

When to use: End of session, before committing.

What it does:

  1. Reviews git diff to see what changed
  2. Summarizes work completed
  3. Identifies what's in progress
  4. Suggests next steps
  5. Outputs formatted summary for STATUS.md

Usage:

code
User: /memory-summary

Workflow Examples

Starting AI Friday Session

code
User: /memory-start

Claude: Reading memory files...

📍 Current State (from STATUS.md):
- Phase: Milestone 1 - Core Referral Tracking
- Task: Implementing referral hash generation (80% complete)
- Last updated: 2025-01-13 16:30

🎯 Next Steps (from ROADMAP.md):
- [ ] Complete hash generation
- [ ] Add cookie tracking
- [ ] Implement conversion detection

📋 Recent Decisions:
- Decision 001: Using SHA-256 for referral hashes

Ready to continue! Let's complete the hash generation.

Ending Session

code
User: /memory-update

Claude: What did you complete this session?
User: Finished hash generation and started cookie tracking

Claude: What's currently in progress? (with %)
User: Cookie tracking is about 40% done

Claude: Any blockers?
User: None

Claude: What should we do next session?
User: Complete cookie tracking and test with MemberPress

Claude: ✅ Updated STATUS.md:

**Last Updated:** 2025-01-20 11:00
**Current Task:** Implementing cookie tracking (40% complete)

## Recent Progress
- ✅ Completed referral hash generation
- ⏳ Started cookie tracking (40%)

## Next Steps
1. Complete cookie tracking implementation
2. Test integration with MemberPress hooks
3. Begin conversion detection

Logging a Decision

code
User: /memory-decision

Claude: What decision was made?
User: Use SHA-256 hashing for referral links instead of sequential IDs

Claude: What alternatives did you consider?
User: 1. Sequential IDs - simple but predictable
      2. UUID - unique but too long for URLs
      3. SHA-256 hash - chosen

Claude: Why did you choose SHA-256?
User: Non-reversible, secure, consistent length, no database lookup needed

Claude: What are the tradeoffs?
User: Pros: Security, no PII in URL
      Cons: Can't reverse-lookup referrer without DB query

Claude: ✅ Added Decision 002 to DECISIONS.md

Best Practices

DO:

  • ✅ Run /memory-start at the beginning of every session
  • ✅ Run /memory-update at the end of every session
  • ✅ Use /memory-decision for significant choices (architecture, tech stack)
  • ✅ Use /memory-learn for non-obvious discoveries
  • ✅ Keep STATUS.md focused on current state only

DON'T:

  • ❌ Skip updating STATUS.md (next session will be confused)
  • ❌ Log trivial decisions (variable names, formatting)
  • ❌ Copy entire codebase into JOURNAL.md (keep entries concise)
  • ❌ Let STATUS.md get stale (update it!)

Integration with AI Friday Workflow

Weekly Cycle:

code
Friday 9:00am  - /memory-start (catch up from last week)
Friday 9:05am  - Start development
Friday 10:45am - /memory-decision (if made architectural choice)
Friday 10:50am - /memory-learn (if discovered something)
Friday 11:00am - /memory-update (capture progress)
Friday 11:05am - Commit changes (including memory files)

Git Workflow:

bash
# At end of session
git add .
git status  # Verify memory files are included

git commit -m "feat: implement cookie tracking

Updated STATUS.md with progress and next steps.

🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>"

Team Adoption

For multi-team projects:

  • Each team branch has own STATUS.md, ROADMAP.md, DECISIONS.md, JOURNAL.md
  • Share DECISIONS.md across teams in presentation
  • Compare ROADMAP.md to see relative progress
  • Learn from each team's JOURNAL.md discoveries

Example:

code
memberpress-referral-program/
├── yellow-team-01/
│   ├── STATUS.md     # Yellow team's current state
│   ├── ROADMAP.md    # Yellow team's tasks
│   └── ...
├── purple-team-01/
│   ├── STATUS.md     # Purple team's current state
│   ├── ROADMAP.md    # Purple team's tasks
│   └── ...

Technical Notes

File Locations:

  • Root level: STATUS.md, ROADMAP.md, DECISIONS.md, JOURNAL.md, CLAUDE.md
  • Domain-specific: [directory]/CLAUDE.md (optional)

Formats:

  • All files are GitHub-flavored markdown
  • Use consistent checkbox syntax: - [x], - ⏳, - [ ]
  • Use ISO date format: YYYY-MM-DD HH:MM
  • Use semantic status indicators: ✅ ⏳ ❌ ⚠️

Git:

  • Memory files SHOULD be committed with code changes
  • Include in .gitignore: Nothing (all memory files should be versioned)
  • Merge conflicts: Most recent wins (timestamps help resolve)

Resources

  • Full documentation: /docs/AI-AGENT-MEMORY-SYSTEM.md
  • Templates: /.claude/skills/ai-memory/templates/
  • Training guide: /.claude/skills/ai-memory/TRAINING.md

Version: 1.0 Last Updated: 2025-01-21 Compatible with: Claude Code, Claude (web/API), ChatGPT, any AI assistant