Working Memory

Quick Reference

• Protocol: CONTINUITY.md read at session start, updated at session end
• Storage: .cm/ directory per project
• Inspired by: Loki Mode (Autonomi)
• Data: Session state + error patterns + architecture decisions

Why Working Memory?

AI agents have amnesia — every new session starts from zero. Without working memory:

• ❌ Same bugs get re-introduced
• ❌ Architecture decisions get forgotten
• ❌ Context switches cause rework
• ❌ Agent doesn't know what it was doing

With Cody Master's working memory:

• ✅ Context persists across sessions
• ✅ Error patterns are captured and prevented
• ✅ Decisions are documented and respected
• ✅ AI picks up exactly where it left off

Setup

# Initialize working memory
cm continuity init

# Check status
cm continuity status

# View learnings
cm continuity learnings

# View decisions
cm continuity decisions

# Reset (careful!)
cm continuity reset

Memory Hierarchy

.cm/
├── CONTINUITY.md      # Working memory (current session)
├── config.yaml        # RARV cycle settings
└── memory/
    ├── learnings.json  # Error patterns (persists forever)
    └── decisions.json  # Architecture decisions (persists forever)

Level	File	Purpose	Lifespan
1	CONTINUITY.md	Current session state	Updated each session
2	learnings.json	Extracted error patterns	Permanent
3	decisions.json	Architecture decisions	Permanent
4	cm-tasks.json	Task queue + RARV logs	Permanent

The Protocol

At Session Start

1. Read .cm/CONTINUITY.md to understand current state
2. Review "Mistakes & Learnings" to avoid past errors
3. Check "Next Actions" to determine what to do
4. Reference Active Goal throughout your work

During Work

PRE-ACT ATTENTION CHECK (before every significant action):
  - Re-read Active Goal
  - Ask: "Does my planned action serve this goal?"
  - Ask: "Am I solving the original problem, not a tangent?"
  - If DRIFT detected → log it → return to goal

At Session End

1. Update "Just Completed" with accomplishments
2. Update "Next Actions" with remaining work
3. Record any new "Mistakes & Learnings"
4. Record any "Key Decisions" made
5. Update "Files Modified" list
6. Set currentPhase and timestamp

On Error

ON_ERROR:
  1. Capture error details (stack trace, context)
  2. Analyze root cause (not just symptoms)
  3. Write learning to CONTINUITY.md
  4. Update approach based on learning
  5. Retry with corrected approach
  6. Max 3 retries per error pattern before ESCALATE

CONTINUITY.md Template

# CodyMaster Working Memory
Last Updated: [ISO timestamp]
Current Phase: [planning|executing|testing|deploying|reviewing]
Current Iteration: [number]
Project: [project name]

## Active Goal
[What we're currently trying to accomplish — 1-2 sentences max]

## Current Task
- ID: [task-id from dashboard]
- Title: [task title]
- Status: [in-progress|blocked|reviewing]
- Skill: [cm-skill being used]
- Started: [timestamp]

## Just Completed
- [Most recent accomplishment with file:line references]
- [Previous accomplishment]

## Next Actions (Priority Order)
1. [Immediate next step]
2. [Following step]
3. [etc]

## Active Blockers
- [Any current blockers]

## Key Decisions This Session
- [Decision]: [Rationale] — [timestamp]

## Mistakes & Learnings
### Pattern: Error → Learning → Prevention
- **What Failed:** [Specific error]
- **Why It Failed:** [Root cause]
- **How to Prevent:** [Concrete prevention action]

Dashboard Integration

Working memory is accessible via the Dashboard API:

# View all projects' memory
GET /api/continuity

# View specific project
GET /api/continuity/:projectId

# Update memory
POST /api/continuity/:projectId

# Add a learning
POST /api/learnings/:projectId
{
  "whatFailed": "Used wrong API endpoint",
  "whyFailed": "Documentation was outdated",
  "howToPrevent": "Always verify endpoints against OpenAPI spec"
}

# Add a decision
POST /api/decisions/:projectId
{
  "decision": "Use Supabase instead of Firebase",
  "rationale": "Better PostgreSQL support, RLS, real-time"
}

Best Practices

✅ Do:

• Read CONTINUITY.md at session start (ALWAYS)
• Update CONTINUITY.md at session end (ALWAYS)
• Record EVERY error in Mistakes & Learnings
• Keep "Just Completed" to last 5 items
• Be specific: "Fixed auth bug in login.ts:42" not "Fixed stuff"

❌ Don't:

• Skip reading CONTINUITY.md ("I remember what I was doing")
• Write vague learnings: "It didn't work" → WHY didn't it work?
• Ignore past learnings when they're relevant
• Let CONTINUITY.md grow beyond ~500 words
• Delete Mistakes & Learnings (archive to learnings.json instead)

Next Steps

• Installation → — First-time setup
• Skills Library → — Browse automation skills
• cm-continuity skill → — Full skill documentation