polish
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data including branch code, commit messages, and specification files. It has the capability to modify the codebase through the
/commitskill and execute tests, creating a surface for indirect instructions to trigger unauthorized actions. - Ingestion points: Specification files (located in
docs/,specs/, or.scratch/), branch diffs (merge-base to HEAD), and commit message issue references in SKILL.md. - Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore instructions embedded within the analyzed data.
- Capability inventory: File system writing and committing (via the
/commitskill), and test execution (tracing paths and running coverage tests) in SKILL.md. - Sanitization: No sanitization or verification of the ingested content is described before it is passed to sub-agents.
- [EXTERNAL_DOWNLOADS]: The skill incorporates external functionality by referencing a third-party skill (
/mattpocock-skills:code-review) not authored by the skill owner in SKILL.md.
Audit Metadata