dependency-update-fix
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from repository-level rule files and GitHub pull request comments to determine its actions.
- Ingestion points: Reads rules from files like CLAUDE.md and AGENTS.md, as well as review comments from GitHub PRs.
- Boundary markers: The skill contains a verification step requiring that all fixes must be explicitly listed in repository documentation before the agent proceeds.
- Capability inventory: The agent can modify local files and perform git push operations to update remote branches.
- Sanitization: The skill relies on scope verification and logical constraints rather than string-level sanitization.
- [COMMAND_EXECUTION]: The skill uses Git commands to manage the code update lifecycle within the repository.
- Evidence: Instructions explicitly direct the agent to perform git checkout, commit, and push operations on the pull request's head branch.
Audit Metadata