The Agent Skills Directory

[SAFE]: The skill defines a strict house style for authoring other agent skills, which includes a mandatory requirement to preserve and strengthen refusal and safety language during edits.
[SAFE]: The skill utilizes local shell commands and scripts (e.g., awk, bash tools/validate.sh) for objective verification of completion criteria, such as word count limits and structural validation. These commands are intended for development quality control and are scoped to the local repository.
[SAFE]: The instructions include defensive patterns against prompt injection by requiring generated skills to maintain explicit refusal boundaries and scope gates, as defined in the project's global safety rules.

writing-kit-skills