vcs-commit-msg
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local version control commands (such as
git logandsvn log) to retrieve repository history, and standard commit commands to apply the final message. These operations are restricted to local history inspection and standard VCS workflows. - [INDIRECT_PROMPT_INJECTION]: The skill reads historical commit messages from the repository, which represents a potential ingestion surface for untrusted data if the repository contains external contributions.
- Ingestion points: History retrieval via
git logorsvn logcommand patterns inSKILL.md. - Boundary markers: Present. The procedure enforces strict structural boundaries, utilizing
--- commit ---block markers along with explicitauthor:,subject:, andbody:fields to isolate data content. - Capability inventory: Limited to running local VCS commands (
git,hg,jj,svn) for logging and committing. - Sanitization: The skill mitigates risks by instructing the agent to only extract stylistic dimensions (such as length, language, and capitalization) and explicitly mandates a human-in-the-loop confirmation step ("Show before applying") before any commit action occurs.
Audit Metadata