1
# Performance Investigation
2
3
Performance work done by intuition optimizes the code that is easy to read rather than
4
the code that is slow. This workflow insists on a baseline before any change, a profile
5
before any theory, one change at a time, and a verification that reports the real
6
number including the cases that got worse.
7
8
## When to use
9
10
- Something got slow and you need to know why, not guess.
11
- Before a rewrite justified by performance, to check the premise.
12
- When an optimization is proposed and nobody has measured the thing it targets.
13
14
## Install
15
16
```bash
17
npx openagents-cli add openagents/perf-investigation
18
```
19
20
| Runtime | Installed to |
21
|---|---|
22
| `claude-code` | `.claude/skills/perf-investigation/` |
23
| `cursor` | `.cursor/rules/perf-investigation/` |
24
| `codex` | `.codex/skills/perf-investigation/` |
25
| `openai-agents` | `agents/perf-investigation/` |
26
| `generic` | `.openagents/perf-investigation/` |
27
28
## What is in the package
29
30
- `WORKFLOW.md` - the loop: baseline, profile, one fix, verify.
31
- `measurement.md` - how to measure without fooling yourself.
32
33
## License
34
35
MIT
36