The Agent Skills Directory

[SAFE]: The skill serves as a defensive methodology for auditing AI agent architectures against established industry standards such as OWASP and NIST. It contains no malicious patterns, persistence mechanisms, or obfuscation.- [PROMPT_INJECTION]: The skill includes a dedicated safety notice that instructs the agent to treat target content as untrusted, preventing the agent from being manipulated by indirect prompt injections present in the files being reviewed.- [COMMAND_EXECUTION]: Tool usage is limited to 'Read', 'Grep', and 'Glob', which are used exclusively for static analysis and pattern matching within the target filesystem. No arbitrary shell command execution is present.- [EXTERNAL_DOWNLOADS]: The skill references the 'fabraix/playground' GitHub repository in its documentation as a tool for manual exploit testing. This is a static reference for the user and does not involve automated or hidden downloads.

agent-security