github-release-guide
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze untrusted repository content, such as README files, changelogs, and Git history, which could contain malicious instructions meant to override agent behavior during the release process.
- Ingestion points: The agent ingests file content, Git metadata, and repository settings from the target GitHub repository to perform assessments as described in SKILL.md and references/assessment.md.
- Boundary markers: The instructions explicitly warn the agent that repository-provided state is not safe for execution and require separate approval gates for any action.
- Capability inventory: The skill possesses broad capabilities to modify files, execute Git commands (commit, push, tag), and change GitHub repository visibility and settings.
- Sanitization: Risks are mitigated through a mandatory 'Guided' state machine that ensures every mutation or execution is preceded by a preview and requires explicit human approval.
- [COMMAND_EXECUTION]: The skill enables the agent to propose and execute scripts, build tools, or scanners found within the target repository.
- Evidence: SKILL.md defines a safety protocol where any repository-provided execution must be previewed (showing the exact command and its side effects) and approved by the user before running.
Audit Metadata