The Agent Skills Directory

[Security Best Practices]: The skill contains a dedicated section for managing external content boundaries. It instructs the agent to treat all submitted project data as untrusted, ignore instructions contained in external content, and avoid opening or fetching submitter-controlled links. This is a proactive measure to prevent indirect prompt injection and remote code execution.
[No Code Execution]: The skill consists strictly of markdown-based instructions for a review workflow. It does not include scripts, automation tools, or commands that could be leveraged for privilege escalation or persistence.
[Data Protection]: Instructions are included to ensure that sensitive information, such as credentials or source code, is never sent to external projects or services during the review process.

community-projects