recall

Searches project memory notes and returns ranked governed context for a query.

20|2|Updated May 16, 2026
One-click install
npx skills add https://github.com/reddb-io/red-skills --skill recall-reddb-io
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: recall
Source: https://github.com/reddb-io/red-skills/tree/main/plugins/memory/skills/core/recall
Command: npx skills add https://github.com/reddb-io/red-skills --skill recall-reddb-io

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? After context resets, agents and developers lose track of past decisions, gotchas, and validated facts, causing repeated mistakes and re-derived work. This Skill searches the project's configured memory surface and returns relevant stored facts ranked by relevance. ## Core Features & Use Cases - Ranked Memory Search: Full-text search in markdown-only mode or a governed graph recall engine that ranks hits by tier, trust, recency, and centrality while hiding superseded nodes. - Graph Read Verbs: In graph mode, exposes search, neighbors, traverse, path, and stats read operations plus MCP, HTTP, and context-pack diagnostics. - Use Case: Before implementing a fix, ask "did we decide how to handle retries?" and recall surfaces the stored decision with its evidence, so you verify and reuse it instead of re-deriving it. ## Quick Start Ask the agent to run memory recall with the query "what do we know about the retry policy" and fold the ranked hits into its answer.

Frequently Asked Questions about recall

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I search project memory for past decisions?▼

Run the recall command with your query terms, for example recall followed by the topic, and optionally add --limit N to cap results. The skill returns ranked hits from stored memory notes that you fold into your answer.

What is the difference between markdown-only and graph mode memory recall?▼

Markdown-only mode full-text-searches stored notes directly. Graph mode runs a governed recall engine that expands deterministic text seeds through the graph neighborhood and ranks results by tier, trust, recency, and centrality.

Why does memory recall return no results?▼

Empty results usually mean memory init has not run or the fact was never stored with the store command. An empty result does not mean the fact is false; it may simply be unstored.

Does memory recall require any setup before use?▼

Yes, memory init must have run to configure the project's memory surface and routing mode. If memory is not configured, recall has nothing to search and will suggest running init first.

Can I trust recalled memory facts without verification?▼

No, treat a recalled note as a claim made at store time and verify stale or high-impact claims before relying on them. Recall provides a governed candidate set, not guaranteed current truth.