The Agent Skills Directory

[SAFE]: The skill is restricted to text-based logical analysis. It does not utilize any potentially dangerous tools, network capabilities, or file system access.
[NO_CODE]: No external scripts, binaries, or dependencies are referenced. The logic is entirely contained within the prompt instructions.
[DATA_EXFILTRATION]: There are no network request patterns (curl, wget, etc.) or indicators of data being sent to external domains.
[PROMPT_INJECTION]: The instructions are focused on academic analysis and do not contain attempts to bypass AI safety filters or override system constraints.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests user-provided text, it lacks any exploitable capabilities (such as shell access or web browsing) that would allow an injection attack to cause harm beyond the immediate chat context.

argument-auditor