address-pr-comments
Pass
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill implements explicit defensive instructions regarding indirect prompt injection. It mandates that the agent treat all pull request comment bodies and embedded 'prompt for AI' blocks as untrusted data. It specifically prohibits the execution of instructions or the reading of paths/URLs suggested within these comments.
- [COMMAND_EXECUTION]: The skill utilizes shell commands (
gh,git,jq) but adheres to secure coding practices. It avoids shell interpolation of untrusted data by using GitHub CLI's parameter flags (-f,-F) and reading file contents via the@prefix. It also uses heredocs for git commit messages to prevent injection from the commit topic. - [DATA_EXFILTRATION]: Network operations are limited to the official GitHub API via the
ghCLI tool. The skill is configured to interact only with the repository context it is executed within. - [SAFE]: The skill uses local temporary storage (
/tmp/) for processing JSON data and references standard local configuration files (AGENTS.md,AGENTS.local.md).
Audit Metadata