ship
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell commands to interact with the local repository, including 'git describe' to find tags, 'git log' to analyze history, and 'grep' to search for hardcoded secrets (API keys, private keys, and authorization tokens).
- [DYNAMIC_EXECUTION]: The skill executes 'verify' and 'release' commands that are dynamically defined in the project's 'AGENTS.md' file. This behavior is central to the skill's purpose of orchestrating project-specific release pipelines.
- [INDIRECT_PROMPT_INJECTION]: The skill represents an indirect injection surface as it parses untrusted data (git commit messages) to determine logic branches (SemVer version bumps).
- Ingestion points: Git commit subjects and bodies via 'git log'.
- Boundary markers: None specified for the commit log parsing.
- Capability inventory: Shell execution for project verification and release tagging.
- Sanitization: The skill uses specific string matching patterns (e.g., 'feat:', 'BREAKING CHANGE') to filter the input data, reducing the likelihood of accidental obedience to instructions embedded in commit messages.
Audit Metadata