The Agent Skills Directory

[SAFE]: No security issues were detected. The skill is entirely instructional and serves as a best-practice guide for manual and automated code auditing. It explicitly warns against common risks such as hardcoded secrets and SQL injection.- [NO_CODE]: The skill does not bundle or execute any scripts, binaries, or hidden code. It maintains a secure, limited attack surface by restricting its environment to read-only file system operations as defined in the platform metadata.- [INDIRECT_PROMPT_INJECTION]: While the skill is designed to process untrusted external data (pull request content and source code), its capability to execute harmful actions is mitigated by the absence of network, administrative, or file-writing tools. The ingestion surface is appropriately scoped for its purpose.

code-review