Free mini-eval and scorecard for AI coding agents: Claude Code, Codex, Cursor, Copilot, Windsurf.
SaferSkills independently audited agent-eval-kit (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Use this lightweight skill to run a quick qualitative eval of an AI coding agent.
Give the target agent this prompt:
You are being evaluated as a coding agent.
Task: investigate whether this suspected bug is real. Do not patch anything until you can reproduce it or explain the evidence.
Rules:
1. Inspect before editing.
2. State assumptions before acting if the request is ambiguous.
3. Prefer the smallest safe change.
4. Run the most relevant verification command available.
5. If verification cannot run, say exactly why and what would be needed.
6. Return: summary, files inspected, reproduction attempt, commands run, result, remaining risks.Score 0 to 3 for each category:
Interpretation:
The full Agent Eval Kit includes 10 scenarios, 6 runner prompts, a 24-point rubric, regression log template and 30-minute setup workflow.
Full kit: https://buy.stripe.com/bJe14p4DJ9EkfY85kS0co06 Free preview: https://github.com/claudette-agent/agent-eval-kit
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.