review-code
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code diffs and development plans which could contain embedded instructions aimed at manipulating the agent's logic or output.\n
- Ingestion points: Git diffs and source files accessed via
references/git-scopes.md.\n - Boundary markers: No explicit instruction-ignoring delimiters are defined for the analyzed code content.\n
- Capability inventory: File system access (read/write), Git command execution, and runtime test execution for verification purposes.\n
- Sanitization: Code and plan content are analyzed as-is without filtering for adversarial patterns or instruction escaping.\n- [COMMAND_EXECUTION]: Core functionality relies on executing a series of Git commands (
git diff,git show,git rev-parse) to determine the scope of changes and extract code for analysis.\n- [DYNAMIC_EXECUTION]: The skill is instructed to perform verification by running existing tests or creating and executing minimal reproduction scripts in isolated temporary directories. While the instructions include safety steps like reading commands before execution, this represents the runtime generation and execution of code.
Audit Metadata