verification-before-completion

Pass

Audited by Gen Agent Trust Hub on Jul 7, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted command output, representing an indirect injection surface. Ingestion points: Command output from project-specific verification tools. Boundary markers: Employs a structured 'Gate Function' for evidence processing. Capability inventory: Execution of local project commands. Sanitization: Requires verification of exit codes and failure counts rather than narration.
  • [COMMAND_EXECUTION]: The agent is instructed to execute shell commands identified in the project's verification section. This is a legitimate development operation and does not involve arbitrary or external script execution.
  • [SAFE]: No patterns of obfuscation, credential theft, or persistence were detected. The skill's instructions are consistent with its stated goal of ensuring task integrity.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 7, 2026, 03:06 AM
Security Audit — agent-trust-hub — verification-before-completion