The Agent Skills Directory

[SAFE]: The skill is entirely composed of natural language instructions and a checklist for a "discipline" of numerical verification. It does not perform any automated actions or execute scripts.
[NO_CODE]: There are no scripts, binaries, or command-line instructions included in the skill. It describes a manual or logic-based process for an agent to follow when claiming a result is "done".
[METADATA_POISONING]: The metadata fields (name and description) accurately reflect the content of the skill and do not contain deceptive instructions or injection patterns.

acceptance-gate