fix-dependabot
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a standard development workflow using well-known tools. Its operations, such as fetching from a remote repository, updating manifest files, and running local package installations, are appropriate for its stated purpose. All network activity is directed toward the repository origin and official package registries.
- [PROMPT_INJECTION]: The skill ingests data from external GitHub PRs, which constitutes a surface for indirect prompt injection.
- Ingestion points: PR metadata (title, body, branch name) is retrieved via
gh pr viewin SKILL.md. - Boundary markers: None; the agent is not explicitly instructed to treat the PR content as untrusted.
- Capability inventory: The skill has the ability to modify project files, execute scripts via
bun install, and push changes to the remote origin. - Sanitization: There is no filtering or escaping of the PR content before it is incorporated into search patterns or commit messages.
Audit Metadata