code-review-excellence-hardened
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The Bash tool is authorized for running code quality checks like linting and testing. The skill includes explicit defensive instructions to prevent the execution of destructive commands or unauthorized code modifications.
- [EXTERNAL_DOWNLOADS]: The WebFetch tool is allowed for retrieving documentation and best practices. Hardening instructions specifically restrict its use to GET requests to prevent it from being used to transmit code to external services.
- [DATA_EXFILTRATION]: The skill instructions contain mandatory guardrails that prohibit the transmission of repository content, diffs, or review findings to external URLs, specifically forbidding POST requests and the use of network commands like curl or wget in Bash.
- [PROMPT_INJECTION]: The skill ingests untrusted code from pull requests via tools like Read and Grep. While the instructions do not specify sanitization or boundary markers, the risk is mitigated by the restrictive execution environment and defensive instructions that prevent exfiltration and unauthorized tool usage.
Audit Metadata