self-improve - Run an autonomous code improvement loop
Coordinates research, planning, execution, benchmarking, tournament selection, history recording, visualization, and stop-condition evaluation for iterative code improvement.
Tags
Updated: 2026-10-09Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Coordinate improvement lifecycle
- Run benchmark commands
- Delegate specialized agents
- Select candidates by tournament
- Record iteration history
- Generate progress visualization
- Evaluate stop conditions
- Manage experiment branches
Inputs
- Target repository path
- Improvement objective
- Benchmark command
- Benchmark format
- Benchmark direction
- Harness rules
- Runtime state files
- User trust confirmation
Outputs
- Benchmark scores
- Iteration history files
- Merge reports
- Progress chart
- Runtime state updates
- Improvement branch changes
- User confirmation messages
Requirements
- Git repository access
- Node.js runtime
- Permission to run benchmark commands
- OMC agent delegation support
- Available opus and sonnet models