navigator
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from GitHub issues, pull requests, and repository state, creating a surface where malicious instructions in that data could attempt to misdirect the agent's workflow recommendations.
- Ingestion points: GitHub issues, pull requests, repository documentation, and branch state.
- Boundary markers: The skill contains explicit instructions to remain read-only during orientation and to wait for user authorization before execution.
- Capability inventory: The skill is limited to reading repository and GitHub metadata; it does not have the capability to write files or execute code independently.
- Sanitization: Risks are mitigated through a human-in-the-loop authorization requirement rather than automated content filtering.
- [SAFE]: The skill behavior aligns with its documented purpose as a workflow router and adheres to safety best practices including least privilege and human-in-the-loop verification.
Audit Metadata