Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
A self-evolution engine for AI agents that records task outcomes, analyzes failures using an LLM, and stores extracted lessons in a searchable experience library. Agents can query past errors semantically and apply suggested fixes on subsequent attempts. Experience is persisted in SQLite with optional vector indexing via ChromaDB.
Storing and retrieving error-specific solutions semantically means agents can reuse past fixes instead of re-analyzing identical failures from scratch.
When to use it
- Diagnosing why an agent keeps failing the same code generation task
- Searching past error logs for a known ValueError before retrying
- Tracking success rate trends across different task types over time
- Automatically applying a previously discovered fix when a similar error recurs
- Auditing an agent's improvement history after a performance tuning session
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: 'This task keeps failing, analyze what's going wrong'
- 1Run evolution_cli.py analyze with the error output from the failed task
- 2Retrieve similar historical experiences ranked by semantic similarity
- 3Inspect suggested solutions and their past success rates
- 4Apply the highest-ranked solution to the next execution attempt
- 5Log the new outcome to update the experience library
Agent identifies a recurring ValueError, retrieves a prior fix (use absolute value for negative inputs, 95% similarity), applies it, and records the result
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.