The Agent Skills Directory

[SAFE]: The skill serves as a knowledge base and instructional set for performing security audits. It promotes the application of industry-standard security controls and secure coding practices. All instructions aim to improve the security posture of the user's project.
[METADATA_POISONING]: The skill metadata correctly identifies the author and purpose without providing deceptive or misleading instructions.
[PROMPT_INJECTION]: No patterns associated with prompt injection, such as instructions to ignore safety guidelines or bypass system prompts, were found in the skill or its references.
[DATA_EXFILTRATION]: The skill explicitly instructs the agent to avoid requesting, logging, or committing sensitive data such as API keys, passwords, or private keys. No network-based exfiltration patterns were detected.
[REMOTE_CODE_EXECUTION]: The skill does not contain instructions for downloading and executing arbitrary remote scripts. All referenced URLs provide documentation from official and reputable security sources.
[INDIRECT_PROMPT_INJECTION]: While the skill processes project source code, which is technically untrusted data, it treats this input as a target for audit rather than executable instructions. It instructs the agent to apply strict validation and sanitization rules to such data, minimizing the risk of accidental obedience to embedded instructions.

security-best-practices