Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Logs errors, corrections, and feature requests to local markdown files during AI agent sessions. Entries use structured fields for priority, area, and status, and can be promoted to permanent project files like CLAUDE.md or AGENTS.md. Hook scripts enable automatic logging when commands fail or prompts are submitted.
Persistent structured logging lets agents avoid repeating the same mistakes across sessions, unlike in-context memory that resets each time.
When to use it
- Logging a bash command failure mid-task for later review
- Recording a correction when the agent used the wrong package manager
- Capturing a feature request the agent cannot fulfill yet
- Promoting a recurring best practice into the project's CLAUDE.md
- Reviewing past errors before starting a database migration
Example Workflow
Here's how your AI assistant might use this skill in practice.
User says: 'Actually, this project uses pnpm, not npm'
- 1Detect correction trigger from user message
- 2Append entry to .learnings/LEARNINGS.md with category 'correction' and area 'config'
- 3Search existing entries for related package manager issues with grep
- 4Distill the correction into a concise rule and add it to CLAUDE.md under Build & Dependencies
Learning entry logged with a generated ID and status 'pending', correction promoted to CLAUDE.md