pre-merge-audit
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes external data including repository source code and staged git diffs (SKILL.md).
- Ingestion points: The agent reads staged diffs and project files during pre-commit passes and deep cleanups.
- Boundary markers: Absent; the instructions do not explicitly require the use of delimiters or 'ignore' instructions for the code being analyzed.
- Capability inventory: The skill has file-write access to edit code and command execution capabilities to run verification tools like lints, typechecks, and tests (SKILL.md, references/cleanup.md).
- Sanitization: Absent; there are no instructions to sanitize or escape the content of the diffs.
- Mitigation: The skill mitigates this by stating 'the user still owns the commit itself' and requiring a 'decisive check' where a runnable probe must fail to justify a cut.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute local verification tools, including lints, typechecks, and project-specific 'build gates' (SKILL.md, references/cleanup.md). These commands are standard for developer workflows and are used here exclusively to validate the safety of code deletions and restructurings.
Audit Metadata