memory-init

Initialize Project Memory

Set up the .ai-memory/ directory in the current project root for cross-tool decision tracking.

Steps

1. Create the .ai-memory/ directory in the project root (current working directory).

2. Create these files:

   • .ai-memory/decisions.jsonl — empty file (canonical decision store)
   • .ai-memory/research.jsonl — empty file (canonical research store)
   • .ai-memory/metadata.json — with this content:

     {
       "tokenCount": 0,
       "lastSync": null,
       "sessionCount": 0,
       "decisionCount": 0,
       "researchCount": 0,
       "researchTokenCount": 0
     }
     

   • .ai-memory/.gitignore — with this content:

     # Transcript data is ephemeral and potentially sensitive
     .last-session.txt
     

3. Run the sync script to generate tool-specific instruction files: Run via Bash: node -e "require('${CLAUDE_PLUGIN_ROOT}/scripts/sync-tools.js').syncAll(process.cwd())"

   This will create:

   • .github/copilot-instructions.md (with self-referential save instructions for Copilot)
   • .cursor/rules/project-decisions.mdc (with frontmatter + save instructions for Cursor)
   • CLAUDE.md (with project decisions section between markers, or update existing)

4. Report success with a summary:

   • List all created files
   • Suggest: "Add .ai-memory/ to git to share decisions across the team"
   • Suggest: "Use /project-memory:memory-save to record your first decision"
   • Note: "Copilot and Cursor instruction files include self-referential save instructions — those tools can now save decisions directly to the canonical store"