verification-before-completion
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWPROMPT_INJECTIONNO_CODE
Full Analysis
- PROMPT_INJECTION (LOW): The skill utilizes strong imperative and coercive language ('If you lie, you'll be replaced', 'Iron Law', 'Non-negotiable') to enforce behavioral compliance. While used as a process guardrail, these represent instruction-override patterns.
- NO_CODE (INFO): The skill is purely instructional and contains no executable scripts, configurations, or binaries, which significantly reduces the technical attack surface.
- INDIRECT PROMPT INJECTION (LOW): The protocol requires agents to ingest and read the 'Full output' of external commands, which is an ingestion point for untrusted data. Ingestion points: Output from verification commands (tests, linters, builds). Boundary markers: None defined. Capability inventory: Implies command execution and file reading capabilities in the agent context. Sanitization: No instructions provided to sanitize or escape command output before processing.
Audit Metadata