google-gemini / google-gemini/gemini-cli
Investigate Subagent Eval Strategy
Open
🔒 maintainer only
aiq/eval_infra
area/agent
kind/customer-issue
priority/p2
status/bot-triaged
workstream-rollup
- Dominant language
- TypeScript
- Stars
- 107k
- Forks
- 14.6k
- Avg merge
- 2d 3h
- Merged PRs (30d)
- 45
Description
Create an initial set of evals (behavioral, benchmarks, etc) that capture the effectiveness of subagents.
Objectives:
- Find and document initial mechanism to reliably evaluate subagent performance
- Document single use case that is effective and ensure evals pass
Contributor guide
Assessment
This issue has not been assessed yet.