The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or security vulnerabilities were detected across the skill files. The content consists of instructional markdown, report templates, and evaluation scenarios designed to guide an agent through a structured benchmarking process.
[NO_CODE]: The skill manifest and file structure confirm that no executable code (Python, Node.js, or shell scripts) is included. The 'scriptPolicy' is set to 'data-only', and there are no instances of remote code execution patterns or unverifiable dependencies.
[PROMPT_INJECTION]: The instructions do not contain attempts to override agent safety filters or bypass system guidelines. Mentions of 'ignore' are used in a benign instructional context, such as advising against post-hoc metric invention.

fable-eval