google-gemini / google-gemini/gemini-cli
[Task] Establish Behavioral Evals for Verbosity
Open
🔒 maintainer only
area/agent
kind/enhancement
priority/p1
workstream-rollup
- Dominant language
- TypeScript
- Stars
- 107k
- Forks
- 14.6k
- Avg merge
- 2d 3h
- Merged PRs (30d)
- 45
Description
### Establish Behavioral Evals for Verbosity
**Description:**
Create or update evaluation scripts in `evals/` to measure output length and "noise" levels. Use these as a baseline to validate that Phase 1 changes actually reduce verbosity and ensure the agent isn't becoming "too quiet" again.
**Part of:** #21304
Contributor guide
Assessment
This issue has not been assessed yet.