verification-before-completion
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [NO_CODE] (SAFE): The skill is entirely instruction-based and does not include any executable scripts, binaries, or configuration files.
- [PROMPT_INJECTION] (SAFE): The forceful language used (e.g., 'The Iron Law', 'non-negotiable') serves as a behavioral reinforcement technique for quality assurance. It does not attempt to bypass safety filters, extract system prompts, or subvert the AI's core alignment.
- [COMMAND_EXECUTION] (SAFE): While the skill instructs the agent to run verification commands, it provides no hardcoded command strings or malicious payloads, relying on the agent's existing project-specific tools.
- [DATA_EXFILTRATION] (SAFE): No network operations, external domains, or hardcoded credentials were found.
- [INDIRECT_PROMPT_INJECTION] (SAFE): The skill requires the agent to read tool output, creating a potential surface for indirect injection if a tool's output is compromised. However, this is a standard operational requirement for development tasks. 1. Ingestion points: Command/tool outputs (e.g., test results). 2. Boundary markers: None explicitly defined. 3. Capability inventory: Execution of local development tools (tests, linters). 4. Sanitization: None specified.
Audit Metadata