write-conventional-commit
Pass
Audited by Gen Agent Trust Hub on Apr 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill executes local git commands such as
git statusandgit diffto retrieve the state of the repository. These operations access source code contents to generate summaries. No network activity or exfiltration patterns were detected. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted input from the local file system via the
git diffcommand output (SKILL.md). There are no specific boundary markers for this data. The capability inventory is restricted to local git read operations. Mitigation is provided through strict output formatting rules (e.g., "return exactly one final commit message and nothing else") that discourage the agent from following instructions embedded in code changes. No explicit sanitization of the diff content is performed.
Audit Metadata