The Agent Skills Directory

[NO_CODE]: The skill is entirely comprised of Markdown instructions and evaluation guidelines. There are no scripts, binaries, or configuration files that execute commands on the host system.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze external SKILL.md files, which are untrusted data sources.
Ingestion points: Instructions in SKILL.md and README.md direct the agent to "Read SKILL.md completely" from a user-provided file path.
Boundary markers: The instructions do not define specific delimiters or "ignore instructions" wrappers for the content being analyzed, relying on the model's ability to distinguish between its instructions and the data being scored.
Capability inventory: The skill itself has no capabilities (no subprocess calls, no network ops); it relies on the agent's default file-reading tools.
Sanitization: No sanitization or filtering of the input skill content is specified.
[INDIRECT_PROMPT_INJECTION]: The skill contains a self-referential section titled "Self-Evaluation Note" in SKILL.md, where it evaluates its own design against its criteria. This content targets the analysis process itself, though it appears to be for educational/demonstrative purposes rather than an attempt to bypass security filters.

skill-judge