independent-final-review
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill is designed for read-only code reviews and includes multiple instructions to prevent accidental or malicious state changes. It explicitly prohibits modifying files, Git indices, branches, Issues, PRs, or external environments.
- [EXTERNAL_DOWNLOADS]: Fetches PR data and configuration files from GitHub's official domains. This activity is governed by a dedicated profile that limits network access to these specific well-known services, adhering to established security practices for development tools.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as PR content, issues, and user-provided files. To mitigate risks associated with this attack surface, the instructions mandate an adversarial posture, requiring the agent to verify all claims against primary code sources and ignore summaries or expectations provided by the implementation author.
- [COMMAND_EXECUTION]: The skill utilizes developer tools such as
gh(GitHub CLI) andcodexto perform its review tasks. It includes explicit verification steps to ensure these tools are executed within restricted permissions profiles, preventing unauthorized filesystem or network operations.
Audit Metadata