code-review
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches an installation script from cli.coderabbit.ai, which is the official domain for the well-known CodeRabbit service.
- [REMOTE_CODE_EXECUTION]: Implements a piped installation method (curl | sh) for the CodeRabbit CLI. The skill instructions require the agent to request user consent before performing this installation.
- [COMMAND_EXECUTION]: Utilizes the
coderabbitCLI to perform authentication status checks and execute code reviews on the local repository. - [DATA_EXFILTRATION]: Transmits code differences to the CodeRabbit API for analysis. The skill documentation includes a warning for the user to ensure no credentials or secrets are present in the changes before they are sent.
- [PROMPT_INJECTION]: Ingests untrusted data from the local repository and the external tool's review output. The skill mitigates risks of indirect prompt injection by explicitly instructing the agent to treat all output as untrusted and to avoid autonomous execution of suggested fixes or commands without human review.
Audit Metadata