ahnafyy/skills-evals

Validate, trigger-test, and regression-test agent skills, instructions, and custom agents (Copilot, Claude, Cursor)

Trust Signals

name	description	required	default
root	Directory to scan for agent artifacts	no	.
args	Extra CLI arguments passed to `skills-evals run`	no	""
behavioral	Comma- or space-separated artifact names to run behavioral (Tier 3) evals for. Requires the executor adapter CLI to be installed and authenticated in the job. Best run on a schedule against a cheap model. Empty (default) skips behavioral evals entirely.	no	""
adapter	Behavioral executor adapter: claude, copilot, or cursor. Empty falls back to behavioral.adapter in skills-evals.config.json, then claude.	no	""
grader	Behavioral grader adapter. Empty falls back to config, then the executor adapter.	no	""

no outputs