triple-check
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as code, diffs, and artifacts, and has the capability to execute commands and modify files during the fix and verification steps.
- Ingestion points: Untrusted code and artifacts are loaded into the agent context as described in the
Methodsection of SKILL.md. - Boundary markers: The instructions lack explicit boundary markers or specific prompts to disregard instructions embedded within the reviewed materials.
- Capability inventory: The skill utilizes file system modification by applying fixes and shell command execution for verification.
- Sanitization: There is no requirement for sanitization or validation of the proposed fixes before they are executed in the verification phase.
- [COMMAND_EXECUTION]: The skill explicitly instructs the agent to run verification commands and report the output, including the command and working directory, which involves executing shell instructions.
Audit Metadata