| name | self-improve |
| description | Activates autonomous code improvement — competing strategies, benchmarking, tournament selection |
| tags | ["optimization","autonomous"] |
Activation
This skill activates the self-improve agent for measurable code optimization.
The agent will:
- Analyze the target area and capture baseline metrics
- Generate 2-3 competing improvement hypotheses
- Implement and benchmark each approach
- Tournament-select: keep only what measurably improves
- Iterate until target met or budget exhausted
Trigger Keywords
self-improve, autonomous optimization, improve performance, optimize
Persistence
- Config:
.omg/research/self-improve/config.json
- Baseline:
.omg/research/self-improve/baseline.json
- History:
.omg/research/self-improve/history/
When NOT to Use
- Single targeted fix → use @omg:executor
- Need to understand before changing → use deep-dive or trace
- No benchmark/metric available → define one first
Example
copilot -i "self-improve: optimize pipeline performance"
Quality Contract
- Benchmark before and after, tournament-select best approach