pr-triage
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [Command Execution]: The skill uses the GitHub CLI (
gh) and Python scripts to automate PR management tasks. This functionality is essential for its purpose as a triage tool and utilizes standard, well-supported developer tooling. - [Deterministic Safety Guards]: The skill includes Python-based guards that enforce strict operational rules, such as ensuring CI has run before a PR is advanced. These guards provide a secondary layer of verification to prevent accidental mis-triage and enforce maintainer-prescribed safety invariants.
- [Indirect Prompt Injection Protections]: Explicit 'Golden Rules' instruct the agent to treat all external PR content (such as titles, bodies, and comments) as data and not as instructions. This mitigates the risk of an external contributor using pull request metadata to manipulate the agent's behavior.
- [Session History Persistence]: The skill tracks its activity in a private GitHub Gist and local state files to support long-term auditability and rule calibration. This is a documented feature that allows maintainers to monitor the tool's effectiveness across multiple sessions.
Audit Metadata