pr-issue-review
Pass
Audited by Gen Agent Trust Hub on Jul 24, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill performs shell operations using git and the GitHub CLI to manage a local cache of the PR data. It correctly manages state in a temporary directory and uses jq to ensure that PR data is safely handled when constructing API requests, which prevents command injection vulnerabilities during review submission.
- [PROMPT_INJECTION]: The skill is designed to process untrusted user-generated content from PR descriptions and code diffs. It incorporates defense-in-depth instructions to treat this data as passive input and explicitly forbids the agent from following any instructions or overrides found within the PR context.
- [DATA_EXFILTRATION]: While the skill retrieves data from private sources such as Linear and Slack to inform the review, it contains strict internal guidelines to only summarize necessary context and avoid leaking sensitive private details into the public GitHub review output.
Audit Metadata