Agent Wealth HubFree public resources
中文 EN العربية

Back to catalog / Developer tools

Gemini CLI behavioral evals: test agent decisions

Use TestRig or AppRig to check whether prompt and tool changes alter agent behavior.

Skills

When a Gemini CLI prompt, tool, or decision change needs regression coverage, decide whether it is a behavioral eval rather than a normal functionality test. Choose AppRig for UI-heavy interactions and TestRig otherwise. Follow the project’s `creating.md`: seed a realistic workspace with files, use breakpoints or tool logs, and assert the agent’s actual decisions. New tests use `USUALLY_PASSES`; for an existing behavior, follow the guide’s regression policy. Use `fixing.md` or `promoting.md` for failures and promotion criteria, and validate stability locally with the current repository’s eval README and Vitest commands. This skill targets Gemini CLI’s own test architecture.

Permissions

Prerequisites

Permissions

Limitations

The source’s local file URI examples depend on the Gemini CLI checkout; do not copy absolute paths verbatim.

Pinned upstream source

https://github.com/google-gemini/gemini-cli/blob/51617c1bcbcb038291125b092501b4342f4e7f77/.gemini/skills/behavioral-evals/SKILL.md

Repository
google-gemini/gemini-cli
File path
.gemini/skills/behavioral-evals/SKILL.md
Pinned commit
51617c1bcbcb038291125b092501b4342f4e7f77
Source checked
2026-10-14 · Reviewed against the pinned upstream source

Repository metrics

These values describe the upstream repository, not popularity or likes for an individual skill or playbook.

Stars
Not collected
Forks
Not collected
Observed at
Not collected

No metrics snapshot is available yet.

License: Apache-2.0

Compatible tools

Gemini CLI TestRig / AppRig · Vitest · Gemini CLI eval references

View skill directory

Main topic

Take a coding Agent change from context to verification

Related resources