logic-gatekeeper
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is purely instructional and contains no executable scripts or system commands.
- [SAFE]: No network operations or external downloads were detected. References to the author's domain (agile-v.org) are standard metadata.
- [SAFE]: The skill encourages safe AI behavior by requiring the agent to 'Halt and Ask' when constraints are unknown or conflicting, rather than assuming or proceeding with potentially unsafe requirements.
- [SAFE]: No data exfiltration or credential exposure patterns were found. The skill operates on local requirements files without sending data externally.
- [INDIRECT_PROMPT_INJECTION]: The skill reads external requirements from
REQUIREMENTS.md(Ingestion point). While it lacks explicit boundary markers like XML tags, it possesses no dangerous capabilities (no subprocess, network, or exec calls) that could be exploited by malicious content within those requirements. It is limited to reading the file and appending findings. Severity is safe due to lack of capabilities.
Audit Metadata