release-skills
Pass
Audited by Gen Agent Trust Hub on Jul 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on shell commands (git and gh) to perform repository operations, version bumping, and release publishing. These operations modify the local state and interact with remote servers.\n- [COMMAND_EXECUTION]: The skill implements a project hook system via the .releaserc.yml configuration file. It can execute arbitrary commands defined in prepare_artifact and publish_artifact hooks. This allows for flexible build and deployment logic but means the agent will run commands specified in the repository's data files.\n- [PROMPT_INJECTION]: The skill is exposed to indirect prompt injection because it ingests and processes untrusted content from the repository's history, including git commit messages and pull request metadata. This information is used to automatically generate changelogs and recommend version increments. \n
- Ingestion points: Commit messages via git log, PR author/details via gh pr view, and local CHANGELOG*.md files.\n
- Boundary markers: None explicitly implemented for parsing commit logs.\n
- Capability inventory: Shell command execution (git/gh), file system write access for version updates and changelog generation, and custom command execution via config hooks.\n
- Sanitization: The skill follows a best practice of writing release notes to temporary files rather than inlining them directly into shell commands, reducing the risk of command injection from malicious commit messages.
Audit Metadata