verification-before-completion

Pass

Audited by Gen Agent Trust Hub on May 6, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is entirely instructional, focusing on a workflow for software development verification (testing, building, and requirements checking).
  • [PROMPT_INJECTION]: While the skill uses strong, mandatory language (e.g., "The Iron Law," "Non-negotiable") to define agent behavior, these instructions are aimed at preventing performance failures and hallucinations rather than bypassing system security or safety filters.
  • [DATA_EXFILTRATION]: No sensitive file paths, hardcoded credentials, or network exfiltration patterns were detected. The skill focuses on local command verification.
  • [REMOTE_CODE_EXECUTION]: The skill mentions running commands like tests or builds as part of a verification workflow, but it does not specify or download any external code, scripts, or unverifiable packages.
Audit Metadata
Risk Level
SAFE
Analyzed
May 6, 2026, 11:48 PM