verification-loop

Pass

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it parses and acts upon data from external tool outputs and git diffs. ● Ingestion points: Output from build, lint, and test tools in SKILL.md. ● Boundary markers: No specific delimiters or warnings are used to distinguish tool output from instructions. ● Capability inventory: The skill can execute various shell commands through package managers and build tools in SKILL.md. ● Sanitization: No sanitization or validation of tool output is performed.
  • [COMMAND_EXECUTION]: The skill's primary function involves running local system commands to verify code quality.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 9, 2026, 10:10 PM