What problem does it solve?
Connecting a memory tool or MCP memory server does not make an agent actually use it. Without standing rules, agents forget user preferences, project conventions, and past decisions between sessions, forcing users to repeat themselves and re-explain context.
Core Features & Use Cases
- Recall-before-acting rules: Defines exactly when to search memory (returning to a project, choosing libraries, writing commits) and when not to (one-off factual questions), avoiding wasted tool calls.
- Save-after-deciding rules: Specifies what deserves a memory entry (decisions, corrections, failures, environment facts) and what must never be saved (secrets, transient state, re-readable file contents).
- Durable memory hygiene: Enforces one-fact-per-entry, closing superseded entries instead of deleting them, and surfacing contradictions rather than silently resolving them.
- Use Case: A user says "stop using npm here, we're on pnpm." The skill directs the agent to save that correction with its reason and date, then recall it before running any package command in future sessions.
Quick Start
Ask the agent to set up persistent memory discipline for this project so it recalls past decisions before acting and saves corrections and conventions after each session.
