Behavioral Evals

This is for when you need to test whether your AI agent is making the right decisions, not just whether your code works. You'll use evalTest for logic validation and appEvalTest for UI-heavy scenarios. The workflow is structured around a policy system: new tests start as USUALLY_PASSES, then get promoted to ALWAYS_PASSES once stable to lock in regression coverage. What I like here is the pragmatism. It includes specific guides for the three things you actually do: creating tests with workspace seeding and breakpoint assertions, fixing failures with investigation steps, and promoting candidates when they're ready. The decision tree at the top saves you from overthinking which approach to use.

skills.sh installs: 294
GitHub stars: 108.7K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add google-gemini/gemini-cli --skill behavioral-evals --agent claude-code

Installs into .claude/skills of the current project.

Files

Behavioral Evals

Overview

Behavioral evaluations (evals) are tests that validate the agent's decision-making (e.g., tool choice) rather than pure functionality. They are critical for verifying prompt changes, debugging steerability, and preventing regressions.

[!NOTE] Single Source of Truth: For core concepts, policies, running tests, and general best practices, always refer to evals/README.md.

🔄 Workflow Decision Tree

Does a prompt/tool change need validation?
- No -> Normal integration tests.
- Yes -> Continue below.
Is it UI/Interaction heavy?
- Yes -> Use appEvalTest (AppRig). See creating.md.
- No -> Use evalTest (TestRig). See creating.md.
Is it a new test?
- Yes -> Set policy to USUALLY_PASSES.
- No -> ALWAYS_PASSES (locks in regression).
Are you fixing a failure or promoting a test?
- Fixing -> See fixing.md.
- Promoting -> See promoting.md.