baekenough/memory-recall

Memory Recall Skill

Search and recall relevant memories from claude-mem using semantic search.

Parameters

| Name | Type | Required | Description | |------|------|----------|-------------| | query | string | yes* | Search query (*not required with --recent) |

Options

--recent, -r     Recall most recent memories
--limit, -l      Maximum results to return (default: 5)
--verbose, -v    Show full memory content
--date, -d       Filter by date (YYYY-MM-DD)

Workflow

1. Build semantic query
   ├── Prefix with "my-project"
   ├── Add user query terms
   └── Include date if specified

2. Search claude-mem
   └── chroma_query_documents

3. Format results
   ├── Sort by relevance score
   ├── Show summary for each
   └── Provide memory IDs for detailed access

4. Present to user

Query Tips

effective_queries:
  - "authentication flow" (topic-based)
  - "2025-01-20 oauth" (temporal + topic)
  - "decision jwt" (type + topic)
  - "agent creation workflow" (semantic intent)

avoid:
  - Single generic words like "code" or "task"
  - Very long queries (keep it focused)
  - Special characters or complex syntax

Retrieval Strategy

Recall-Precision Tradeoff

Default bias: recall > precision — it is easier to filter out irrelevant results (false positives) than to recover missed information (false negatives).

| Task Type | Bias | Recommended --limit | Rationale | |-----------|------|--------------------:|-----------| | Debugging / Investigation | High recall (16:1) | 10-15 | Cast wide net, prune later | | Decision reference | Balanced | 5 (default) | Moderate breadth with manageable noise | | Specific fact lookup | High precision | 3 | Narrow, targeted retrieval |

Guidelines

• Over-retrieve, then filter: When uncertain, request more results and discard irrelevant ones in post-processing
• Narrow progressively: Start broad, refine query only if results are noisy — avoid starting too narrow
• Combine temporal + semantic: Add date filters (--date) to semantic queries for better precision without sacrificing recall

Output Format

Basic Search

[sys-memory-keeper:recall authentication]

Searching memories for: "my-project authentication"

Found 3 relevant memories:

[1] mem_abc123 (Score: 0.95)
    Date: 2025-01-20
    Summary: Implemented OAuth flow with Google provider
    Tags: [authentication, oauth, google]

[2] mem_def456 (Score: 0.87)
    Date: 2025-01-18
    Summary: JWT token decision - RS256 algorithm
    Tags: [authentication, decision, jwt]

[3] mem_ghi789 (Score: 0.82)
    Date: 2025-01-15
    Summary: Authentication architecture discussion
    Tags: [authentication, architecture, planning]

Use "sys-memory-keeper:recall --verbose" for full content.

Verbose Output

[sys-memory-keeper:recall authentication --verbose]

Searching memories for: "my-project authentication"

[1] mem_abc123 (Score: 0.95)
    Date: 2025-01-20
    Tags: [authentication, oauth, google]

    Content:
    ## Session Summary
    Date: 2025-01-20

    ### Tasks Completed
    - Implemented OAuth flow with Google provider
    - Added callback handler for OAuth response
    - Created user session management

    ### Decisions Made
    - Use passport.js for OAuth handling
      Rationale: Well-maintained, good documentation

    ---

[2] mem_def456 (Score: 0.87)
    ...

Recent Memories

[sys-memory-keeper:recall --recent]

Fetching recent memories for: my-project

[1] mem_xyz999 (Score: 1.00)
    Date: 2025-01-24
    Summary: Memory system Phase 1 implementation
    Tags: [session, memory, phase1]

[2] mem_xyz888 (Score: 0.98)
    Date: 2025-01-23
    Summary: Agent identification rules update
    Tags: [session, rules, enforcement]

...

No Results

[sys-memory-keeper:recall "nonexistent topic"]

Searching memories for: "my-project nonexistent topic"

No memories found matching your query.

Suggestions:
- Try different keywords
- Use broader search terms
- Check available memories with "sys-memory-keeper:recall --recent"

Related

• memory-save - Save current context