verification-before-completion
Pass
Audited by Gen Agent Trust Hub on Feb 21, 2026
Risk Level: SAFE
Full Analysis
- Procedural Integrity (SAFE): The skill enforces an 'Evidence Before Assertions' principle. It uses strong instructional language to ensure the agent verifies work through fresh evidence. This is a behavioral constraint for quality control and does not attempt to override safety protocols.\n- Command Execution (SAFE): The skill directs the agent to run verification commands like tests or builds. As this is the core intended purpose of the skill and does not involve specific malicious payloads or remote code execution, it is considered safe within context.\n- Indirect Prompt Injection Surface (SAFE): The skill describes processing data such as agent reports and VCS diffs. This is an ingestion point for external data, but the skill explicitly instructs the agent to treat these as untrusted until verified, thereby reducing the risk of indirect injection.\n
- Ingestion points: Agent reports, VCS diffs mentioned in SKILL.md.\n
- Boundary markers: Absent.\n
- Capability inventory: Generic command execution (tests, builds, linters).\n
- Sanitization: Absent.
Audit Metadata