---

name: search description: This skill should be used when the user asks to "search memory", "find in past sessions", "what did we decide about", "look up previous work on", or needs to retrieve specific information from project memory. context: fork hooks: Stop: - hooks: - type: command command: "${CLAUDE_PLUGIN_ROOT}/scripts/validate-search.sh" timeout: 10

Search - Query Project Memory

Search engram semantic memory for past context, decisions, and lessons.

Execution

Step 1: Parse Search Intent

From $ARGUMENTS, identify:

• Query: What the user wants to find
• Scope: Project-only or cross-project
• Type: General search or insights (decisions/lessons)

Step 2: Select Search Strategy

Step 3: Execute Search

General search:

Copymcp__plugin_engram-mcp_engram__memory_search
  query: "{semantic query}"
  n_results: 10

Search with type filter:

Copymcp__plugin_engram-mcp_engram__memory_search
  query: "{semantic query}"
  filter_type: "decision"  # or: remembered, exchange, tool_use, lesson
  n_results: 10

Insights search:

Copymcp__plugin_engram-mcp_engram__memory_insights
  query: "{topic}"
  insight_type: "decision"  # or: lesson, all
  n_results: 5

Step 4: Present Results

Organize results by relevance and type:

Copy## Search Results: "{query}"

### Most Relevant

**[{type}] {date}**
{content snippet}
*Relevance: {score}*

---

### Decisions Found
| Date | Decision | Rationale |
|------|----------|-----------|
| {date} | {choice} | {why} |

### Lessons Found
| Date | Lesson | Category |
|------|--------|----------|
| {date} | {content} | {category} |

### Related Exchanges
{Summary of relevant past conversations}

---

{N} results found. Refine search with more specific terms if needed.

Step 5: Offer Follow-up Actions

Based on results, suggest:

• More specific search queries
• Related topics to explore
• Actions to take based on findings

Query Crafting Tips

Effective Queries

Engram uses semantic search. Write queries as natural language:

Good:

• "how did we implement user authentication"
• "decisions about database schema for orders"
• "bugs related to async state updates"

Poor:

• "auth" (too vague)
• "SELECT * FROM" (not semantic)
• "file.ts line 42" (not indexed)

Using Filters

Filter by type to narrow results:

Exact Matching

Wrap query in quotes for exact phrase matching:

Copymcp__plugin_engram-mcp_engram__memory_search
  query: "\"PostgreSQL JSONB\""

Examples

User: "What did we decide about caching?"

Copymcp__plugin_engram-mcp_engram__memory_insights
  query: "caching"
  insight_type: "decision"
  n_results: 5

User: "Find any gotchas about React hooks"

Copymcp__plugin_engram-mcp_engram__memory_insights
  query: "React hooks"
  insight_type: "lesson"
  n_results: 5

No Results?

If search returns empty:

1. Broaden the query - Use more general terms
2. Check memory stats - Run memory_stats to verify index health
3. Try different filters - Remove type filter for broader results
4. Sync memory - Run memory_sync if recent work isn't indexed