Memory Skill Policy

Router Logic

Scenario 1: User asks "Did we ever...", "Have we...", "I remember..."

1. Recall: Call recall(query) with relevant keywords
2. Synthesize: Combine results into a coherent answer
3. Respond: "I found X memories about that..."

Scenario 2: User asks to "remember this", "store this", "don't forget"

1. Analyze: Determine if this is an insight or an episode
2. Store: Call remember_insight() for learnings, log_episode() for actions
3. Confirm: Show what was stored

Scenario 3: User asks "What have you learned?", "Show memories"

1. List: Call list_harvested_knowledge()
2. Format: Organize by domain/category
3. Present: Show structured summary

Scenario 4: End of significant session

1. Consolidate: Call harvest_session_insight(context_summary, files_changed)
2. Store: Extract key learnings and store them

Workflow: Remembering a Solution

User: How did we fix the nixfmt error before?

Claude:
  1. recall("nixfmt error solution")
  2. → Found: "Use 'just fmt' before committing when nixfmt fails"
  3. → "I remember! We solved this by running 'just fmt'..."

Workflow: Storing an Insight

User: Remember that for this project, all commit messages must be in English.

Claude:
  1. remember_insight(content="All commit messages must be in English only", domain="git")
  2. → Stored in semantic memory
  3. → "✅ Got it! I'll remember that commit messages must be in English."

Workflow: Session Harvesting

User: We're done with this feature. Summarize what we did.

Claude:
  1. harvest_session_insight(context_summary="Implemented knowledge skill for context injection", files_changed=["agent/skills/knowledge/..."])
  2. → Stored as session memory
  3. → "✅ Session harvested. Key learnings stored in Hippocampus."

When NOT to Use Memory Skill

• Quick temporary notes → Use scratchpad instead
• Task tracking → Use backlog/issue tracker
• Code snippets → Use proper documentation

Memory vs Knowledge Skill

Best Practices

1. Store actionable insights, not obvious facts
2. Use domains (git, nix, architecture) for organization
3. Log significant episodes with context
4. Harvest at session end for long-term retention