Skip to content

Repository files navigation

skill-eval-loop

skill-eval-loop is a paired behavioral evaluation tool for Agent Skills. It runs the same task with and without a target skill, keeps the execution conditions fixed, and reports whether the skill materially improved the result.

The evaluator supports Hermes, Claude Code, Codex, and Pi. It uses semantic grading, explicit provenance, response-sensitive grader contrasts, and fail-closed runtime attestation.

Install

Install the skill from this repository with Tink:

tink skill add jon-devlapaz/skill-eval-loop --skill skill-eval-loop
tink skill check

The complete interaction contract is in skills/skill-eval-loop/SKILL.md.

Validate an evaluation suite

export SKILL_EVAL_DIR="$PWD/.agents/skills/skill-eval-loop"

python3 "$SKILL_EVAL_DIR/scripts/audit_suite.py" \
  --skill-path /path/to/target-skill

Auditing is static and does not call a model. See the suite schema and evaluation authoring guide for the format and evidence requirements.

Run a paired diagnostic

Start with a dry run. It validates the plan and reports the exact invocation count without creating a run or calling a model.

python3 "$SKILL_EVAL_DIR/scripts/run_skill_eval.py" \
  --skill-path /path/to/target-skill \
  --harness codex \
  --target-model MODEL_ID \
  --grader-model MODEL_ID \
  --dry-run

Only run the live command after reviewing its invocation count and explicitly authorizing provider calls. The default is a one-trial pilot.

See the harness support matrix for the distinction between adapter implementation and live release evidence.

Development

The evaluator uses the Python standard library at runtime. Development checks require Python 3.11 or newer and Ruff.

python3 -m unittest skills/skill-eval-loop/tests/test_skill_eval_loop.py
bash skills/skill-eval-loop/scripts/healthcheck.sh
ruff check tools tests skills/skill-eval-loop/scripts skills/skill-eval-loop/tests

The release identity verifier checks that Git, Tink receipts, and installed payload bytes and executable modes all identify the same commit:

python3 tools/verify_release_identity.py --help

License

MIT

About

Paired behavioral evaluation for Agent Skills

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages