nkootstra/Skill Eval

Evaluate and benchmark agent skills in pull requests with automated testing and suggestions.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Stale
License
None
namedescriptionrequireddefault
providerLLM provider to use: anthropic, openai, litellm, or claude-codenoanthropic
modelModel name (e.g., claude-haiku-4-5-20250414, haiku, sonnet, opus, gpt-4o, or any LiteLLM model)no""
anthropic_api_keyAnthropic API key (required when provider is anthropic)no
openai_api_keyOpenAI API key (required when provider is openai)no
litellm_api_keyAPI key for LiteLLM providerno
litellm_api_baseBase URL for LiteLLM-compatible APIno
claude_code_oauth_tokenClaude Code OAuth token for Pro/Max subscription auth (generate with 'claude setup-token')no
claude_code_pathPath to Claude Code CLI binary (for claude-code provider; auto-installed if not found)no""
config_pathPath to .skill-eval.yml config fileno.skill-eval.yml
skills_pathPath to skills directory (overrides config file)no
github_tokenGitHub token for PR comments and check statusno${{ github.token }}
parallel_evalsNumber of evals to run in parallel (1-10, lower for rate-limited APIs)no3
eval_presetEval preset: smoke (3 trials), reliable (10 trials), or regression (25 trials). Overridden by explicit eval_trials.no
eval_trialsNumber of trials per eval test case for pass@k/pass^k statistical metrics (1-50, default 1 for single-run)no1
min_pass_rateMinimum eval pass rate (0.0-1.0) to consider the run successful. Default 1.0 requires all evals to pass.no
fail_on_errorFail the action if any evaluation errors are foundnotrue
namedescription
resultsJSON string of evaluation results
passedWhether all evaluations passed
summaryHuman-readable summary of results