code-review
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill constructs and executes shell commands such as
git diff <ponto-fixo>...HEADusing a user-supplied reference point. While the skill includes a validation step usinggit rev-parse, the direct interpolation of user input into a shell command remains a potential vector for command injection depending on the platform's tool execution constraints.\n- [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Category 8) because it ingests untrusted content and processes it via sub-agents without isolation.\n - Ingestion points: Content retrieved via
git diff,git log, and various specification files (e.g., fromdocs/,specs/, or an external issue tracker).\n - Boundary markers: The skill fails to define clear boundary markers or provide instructions to the sub-agents to ignore embedded commands within the analyzed data.\n
- Capability inventory: The skill possesses the ability to execute shell commands and invoke the
Agenttool for sub-tasking.\n - Sanitization: No sanitization, filtering, or escaping is performed on the code changes or documentation before they are interpolated into the sub-agent prompts.
Audit Metadata