Back to skill
Skillv1.0.0
ClawScan security
Chef · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
ReviewMay 1, 2026, 5:30 AM
- Verdict
- Review
- Confidence
- high
- Model
- gpt-5.5
- Summary
- Prompt-injection indicators were detected in the submitted artifacts (unicode-control-chars); human review is required before treating this skill as clean.
- Guidance
- This skill appears safe to install as a cooking advice helper. As with any food-related guidance, users should still verify allergy, food-safety, and dietary-restriction advice for their own situation. ClawScan detected prompt-injection indicators (unicode-control-chars), so this skill requires review even though the model response was benign.
- Findings
[pre-scan:unicode-control-chars] expected: A single control-character signal was reported, but the provided SKILL.md content shows only cooking instructions and emoji metadata, with no visible prompt-injection or hidden behavioral instruction.
Review Dimensions
- Purpose & Capability
- okThe skill’s instructions are coherent with its stated cooking-assistance purpose, covering recipes, substitutions, dietary adaptations, troubleshooting, and kitchen safety.
- Instruction Scope
- okThe instructions stay within culinary guidance and do not ask the agent to override user intent, use tools, access private data, or perform unrelated actions.
- Install Mechanism
- okThere is no install spec and no code files; the artifact is instruction-only.
- Credentials
- okNo binaries, environment variables, credentials, config paths, network access, or file access are requested.
- Persistence & Privilege
- okThe artifacts show no persistence mechanism, background activity, privilege escalation, memory storage, or long-running behavior.