Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
PersistentMind is a Python library that stores typed, searchable memories for AI agents across sessions. It supports seven memory types (fact, preference, procedure, correction, context, relationship, reminder) and three scopes (global, project, session). Data is stored locally with no external dependencies or API keys.
Unlike static prompt files or manual re-explanation, it ranks memories by importance and type, surfacing critical corrections first in every injected context block.
When to use it
- Persisting code style corrections so agents stop repeating the same mistakes
- Storing project-specific commands like migration or deployment steps
- Injecting accumulated project context into new chat sessions automatically
- Building a shared team knowledge base via JSON export and import
- Setting temporary notes that auto-expire after a specified number of days
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Remember that we never use wildcard imports and that migrations run with 'poetry run alembic upgrade head'
- 1Call remember() with the wildcard import rule as MemoryType.CORRECTION, scope PROJECT, importance 10.0
- 2Call remember() with the migration command as MemoryType.PROCEDURE, scope PROJECT
- 3In the next session, call get_context(project='my-app') to retrieve stored memories
- 4Prepend the formatted context block to the incoming user prompt
Context block with the correction surfaced first, followed by the procedure, automatically prepended to every subsequent prompt