The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface because it is designed to ingest and process the content of external, untrusted skill files for evaluation.
Ingestion points: The skill reads SKILL.md and related files from a target directory provided by the user.
Boundary markers: There are no specific instructions or delimiters provided to the agent to distinguish between the skill's own instructions and potential adversarial instructions contained within the analyzed files.
Capability inventory: The skill's capabilities are limited to listing directories, reading files, and writing the final evaluation report to a local path.
Sanitization: The instructions do not include any steps for sanitizing or validating the input text from the audited skills before the evaluation process.
[SAFE]: The skill does not perform any network operations, use hardcoded credentials, or employ obfuscation techniques. All behavior is transparent and aligned with its stated utility.
[NO_CODE]: The provided skill files contain only markdown instructions and reference text. No executable scripts, binaries, or third-party dependencies were found or requested.

skill-evaluation