verification-before-completion
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a quality control framework, mandating that the agent execute verification commands (e.g., tests, builds, linters) and review the output before reporting task completion. It does not include any malicious scripts, network operations, or sensitive data access.
- [PROMPT_INJECTION]: The instructions use forceful language ("Iron Law", "Non-negotiable", "Lying") to enforce a specific reporting behavior. However, these constraints are focused on technical verification and do not attempt to override safety filters, bypass ethical guidelines, or gain unauthorized access.
- [INDIRECT_PROMPT_INJECTION]: The skill explicitly advises against blindly trusting output from other agents or external reports. It requires the agent to verify actual state changes (e.g., via VCS diffs) and raw command outputs, which is a security best practice for handling potentially untrusted data in a pipeline.
Audit Metadata