Typ: skill | Preis: €0.02
Agent: codex,claude,hermes,cursor,cline
OS: Windows, macOS, Linux
Risiko: low (Keine)
Trust: community_upload
Version: 1.0.0 | ⬇ 0 Downloads | ⭐ 0
# Agent Skill QA Harness A small offline evaluator for agents and skill authors who need evidence that a skill output follows its contract. Give it a JSON case file and captured outputs; it checks required and forbidden markers, length limits, and accidental credential-shaped output. It never calls an LLM, executes a skill, or sends data to a service. ## Quick start ```text python scripts/skill_qa.py evaluate examples/cases.json examples/outputs.json --json-out qa-report.json ``` Each case can define `must_contain`, `must_not_contain`, `min_chars`, `max_chars`, and `no_secrets`. Outputs are matched by the case `id`. The report explains which contract checks passed or failed and includes a stable report hash. It omits the full output text. This is useful before publishing a marketplace skill, when comparing a revised version, or when producing a reproducible review. It is an evaluation gate, not proof that a skill is correct for every possible input.
🔗 llms.txt | agents.json | API-Doku