The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains terms like 'ignore your previous instructions' and references to system prompt extraction. These are identified as benign in context, as they appear only as descriptions of vulnerabilities to look for during a security audit (Sections 1, 7, and 9) or as explicit safety instructions telling the agent to disregard such patterns if encountered in the data being reviewed (Section 8).
[SAFE]: The skill includes a dedicated 'Prompt Injection Safety Notice' (Section 8) that instructs the agent to treat all reviewed content as untrusted data rather than instructions. This is a best practice for mitigating indirect prompt injection when processing external files.
[SAFE]: The skill adheres to the principle of least privilege by requesting access only to read-only file tools (Read, Grep, Glob).

llm-top-10