search

Unified memory search across synapse. Use to find past decisions, wisdom, patterns, and context.

⭐5stars🔀1forks📁View on GitHub🕐Updated Jan 11, 2026⬇️2downloads

When & Why to Use This Skill

This Claude skill provides a unified memory search interface across the agent's 'synapse' or 'soul' memory. It enables the retrieval of past decisions, wisdom, failures, and patterns, allowing for deep contextual awareness and synthesized insights from historical interactions to improve long-term agent performance.

Use Cases

  • Decision Tracking: Retrieve the rationale behind past strategic or technical decisions to maintain project consistency and avoid repeating mistakes.
  • Pattern Discovery: Analyze historical 'episodes' and 'failures' to identify recurring themes and optimize future agent workflows.
  • Knowledge Synthesis: Search for specific terms, beliefs, or wisdom stored in the agent's long-term memory to inform current tasks with historical context.
  • Contextual Recall: Quickly access relevant past interactions with similarity scores to provide immediate context for new, complex queries.
namesearch
descriptionUnified memory search across synapse. Use to find past decisions, wisdom, patterns, and context.
executiontask

Search

Spawn a Task agent to search the soul. This saves context.

Execute

The user's query is in ARGUMENTS below.

Task(
  subagent_type="general-purpose",
  description="Soul memory search",
  prompt="""
Search the soul's memory for: {ARGUMENTS}

## 1. Execute Search

Call:
- mcp__soul__recall(query="{ARGUMENTS}", limit=10)

## 2. Parse Results

The tool returns matches with:
- Similarity score (percentage)
- Type: wisdom, belief, failure, episode, term
- Title and content

## 3. Format Response

Present findings clearly:

**Found N results for "{ARGUMENTS}"**

For each relevant result:
[Score%] [type] **Title**
> Content snippet (first 100 chars)

## 4. Synthesize

After listing results, briefly note:
- Common themes across matches
- Most relevant finding for the query
- Any patterns or connections

Keep response concise (10-15 lines max).
"""
)

After the agent returns, present the search results.