ten-principles
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted local file content and Git diff outputs, creating a vulnerability surface where malicious instructions could be embedded in code to influence the agent's review or editing behavior.\n
- Ingestion points: The skill reads output from
git diffand the full content of changed files as defined in the 'Review mode' section.\n - Boundary markers: No explicit delimiters or instructions are provided to separate data (code) from instructions during processing.\n
- Capability inventory: The skill is capable of performing file edits based on findings generated during the review process.\n
- Sanitization: The skill requires explicit user approval for each finding before applying edits, which serves as a human-in-the-loop control to prevent automated exploitation.\n- [COMMAND_EXECUTION]: The skill utilizes system commands to retrieve file differences.\n
- Evidence: The 'Scope' section specifies the execution of
git diff HEADandgit diff <base>to identify changes for analysis.
Audit Metadata