Search persistent memory for context relevant to the current question.
npx skills add https://github.com/hanh-nd/agent-kit --skill recall-hanh-nd
Please help me install this Agent Skill. Skill: recall Source: https://github.com/hanh-nd/agent-kit/tree/main/plugins/memory-kit/.gemini/skills/recall Command: npx skills add https://github.com/hanh-nd/agent-kit --skill recall-hanh-nd
kit_memory_search with concise keywords to find factual memories, keeping names, ticket IDs, and technical terms intact. - Recency Queries: Calls kit_memory_recent for temporal questions like "what did we do last session?", with optional count and source-type filters (digest, concept, entity, preference, wiki). - Hallucination Guard: Returns "No relevant memories found." when nothing matches, never inventing context. - Use Case: A developer asks "what did we decide about the caching layer last week?" and receives a concise summary citing the source file and date. ## Quick Start Ask the agent what was decided or worked on in a previous session and it will search persistent memory for the answer.High-intent search queries and answers about installing and using this skill.
Ask a natural question about prior work, such as what was decided or done last session. The recall skill converts your request into concise keywords and queries persistent memory using kit_memory_search, returning a summary with source and date.
Ask a temporal question like "what did we do last session?" and the skill calls kit_memory_recent. You can specify a count for how many entries to return or narrow results to a source type such as digest, concept, entity, preference, or wiki.
The skill responds with "No relevant memories found." and stops. It does not hallucinate, guess, or perform its own search outside the memory store, so answers stay grounded in actual stored context.
Yes, but only for recency queries via kit_memory_recent. You can explicitly narrow results to source types like digest, concept, entity, preference, or wiki. General semantic searches through kit_memory_search do not use source-type filters.
Results depend on the keywords extracted from your question. The skill strips stop words but keeps names, ticket IDs, and technical terms, so including specific project terms or feature names in your question improves relevance scoring.
ncz-osStore and retrieve persistent memories across agent sessions via API.
zhangloveyanLoad current task progress and resume development after session interruptions.
redisAnswer project questions with citations and ingest sources into structured wiki pages.
sky770825Integrate LanceDB, Git-Notes, and file search for persistent agent memory.