Agent-Skills.md

Agent Skills: Memory Skill Policy

The Hippocampus Interface - Vector-based Memory for LLM (LanceDB + FastEmbed)

UncategorizedID: tao3k/omni-dev-fusion/memory

Repository

tao3k/omni-dev-fusion
tao3kLicense: Apache-2.0
7

Install this agent skill to your local

pnpm dlx add-skill https://github.com/tao3k/omni-dev-fusion/tree/HEAD/assets/skills/memory

Skill Files

Browse the full folder contents for memory.

Download Skill

Loading file tree…

assets/skills/memory/SKILL.md

Skill Metadata

Name
memory
Description
Use for short-term operational memory: track transient issues/workarounds, recall recent execution context, and support cleanup of stale entries.

Memory Skill Policy

This skill is an MCP-facing facade for memory operations. Core memory policy (lifecycle/revalidation/promotion) belongs to Rust memory core, not this skill surface.

Router Logic

Scenario 1: User wants to store a transient finding

  1. Analyze: Determine if the item is transient operational memory (bug/workaround/incident note)
  2. Store: Call save_memory(content, metadata)
  3. Confirm: Show the saved memory ID

Scenario 2: User wants to remember/search

  1. Search: Call search_memory(query, limit)
  2. Format: Present results with relevance scores
  3. Respond: "I found X memories about that..."

Scenario 3: User asks for current operational memory status

  1. List: Call get_memory_stats()
  2. Recall: Call search_memory() with relevant keywords
  3. Present: Show structured summary with transient scope

Commands Reference

| Command | Description | Example | | ------------------ | ---------------------------------------- | -------------------------------------------------------------------------------- | | save_memory | Store short-term operational memory item | save_memory("Temporary workaround for timeout in parser", {"tag": "incident"}) | | search_memory | Semantic search in memory | search_memory("git commit format", limit=5) | | index_memory | Optimize vector index (IVF-FLAT) | index_memory() | | get_memory_stats | Get memory count | get_memory_stats() | | load_skill | Load skill manifest into memory | load_skill("git") |

Workflow: Store a Transient Workaround

User: Remember this temporary fix: increase parser timeout when MCP queue spikes.

Claude:
  1. save_memory(
       content="Temporary workaround: increase parser timeout under MCP queue spike",
       metadata={"domain": "runtime", "kind": "workaround", "source": "user"}
     )
  2. → Saved memory [a1b2c3d4]: Temporary workaround: increase parser timeout...
  3. → "Stored as short-term operational memory."

Workflow: Recall Recent Operational Context

User: What workaround did we use for MCP queue timeout?

Claude:
  1. search_memory("MCP queue timeout workaround")
  2. → Found 2 matches:
     - [Score: 0.8921] Temporary workaround: increase parser timeout...
     - [Score: 0.7234] Queue backpressure note...
  3. → "I found recent operational memory for this issue..."

Memory vs Knowledge Skill

| Aspect | Memory (this skill) | Knowledge skill | | ------------ | ---------------------------------------------- | --------------------------------------------- | | Scope | Short-term operational context | Long-term reusable knowledge | | Nature | Transient (can be purged after revalidation) | Stable (promoted/curated) | | Purpose | "What recent issue/workaround context exists?" | "What durable rule/pattern should be reused?" | | Policy | Managed by Rust memory core lifecycle | Managed by knowledge ingestion/curation flows | | Exposure | MCP tool facade | MCP tool interface |

Best Practices

  1. Store transient operational memory, not permanent rules
  2. Include kind in metadata (incident, workaround, observation)
  3. Use clear, searchable phrasing in content
  4. Promote only proven durable patterns to knowledge