Search Memory

by nowledge-co1f4d7eb61466dbfc93ce8d7a0a806c4d4a196dd0No licenseListed Oct 9, 2026Updated Oct 9, 2026

Search Nowledge Mem memories and threads from Proma when the user references past work, prior decisions, recurring problems, or project history.

Instructions onlyAI & Agents

AI-generated overview

Searches Nowledge Mem memories and threads to surface past decisions, preferences and project history when a user references prior work.

What it does: Guides the agent to proactively query a Nowledge Mem knowledge base when past insights would improve a response. It routes searches first to the MCP memory search tool, then to the thread search tool, and falls back to the nmem CLI with a deep search mode. It also tells the agent to combine nmem results with Proma's built-in recall tool and to frame retrieved context naturally.
When to use it: Use when the user refers to earlier discussions, decisions, preferences or recurring problems, or mentions a project or topic with known history. Also useful when the current problem resembles one solved before. If nothing is found, the skill directs the agent to say so rather than invent context.
Requirements: Requires access to the Nowledge Mem MCP tools (mcpnowledge-memsearch_memories, mcpnowledge-memsearch_threads) or the nmem CLI as a fallback; also references Proma's built-in mcpmemrecall_memory. No scripts are shipped; instructions only.

Search your Nowledge Mem knowledge base proactively when past insights would improve the response.

Strong signals — search without waiting to be asked:

User references "last time", "before", "we discussed", "the other day"
User asks about a decision, preference, or plan that may have been decided before
User mentions a project, repo, or topic with known history
Current problem resembles a previously solved one

Search routing:

Primary (MCP):

mcp__nowledge-mem__search_memories  query: "<search terms>"  limit: 5
mcp__nowledge-mem__search_threads  query: "<search terms>"  limit: 5

mcp__nowledge-mem__search_memories  query: "<search terms>"  limit: 5
mcp__nowledge-mem__search_threads  query: "<search terms>"  limit: 5

Fallback (CLI):

bash

nmem --json m search "auth token rotation" --mode deepnmem --json t search "deploy pipeline discussion" --limit 5

nmem --json m search "auth token rotation" --mode deepnmem --json t search "deploy pipeline discussion" --limit 5

Search both nmem and Proma's built-in mcp__mem__recall_memory — nmem for cross-session history, built-in for current-session notes
Prefer nmem results for decisions, preferences, and project-level context
Frame results naturally — "we decided X last time" not "search result 3 indicates X"
If search returns nothing, say so honestly; don't fabricate context

Source and attribution

License: No license

Content belongs to its original authors. SourceWeft indexes it from a public repository.