The Agent Skills Directory

[SAFE]: The skill acts as a style guide and template for authoring other agent skills. It defines structural rules, word budgets, and canonical one-liners for consistent output. It emphasizes the importance of safety language and refusal boundaries in generated skills.
[COMMAND_EXECUTION]: The instructions include running local validation tools and a word-count command. Specifically, it references bash tools/validate.sh and an awk script to check the body of SKILL.md files. These commands are localized to the repository's internal tools and the files being edited, presenting no evidence of malicious intent or external command injection.
[INDIRECT_PROMPT_INJECTION]: While the skill defines how an agent should process inputs when creating or editing other skills, it does not introduce specific vulnerable surfaces. It provides explicit guidance to maintain safety properties and scope gates in all authored skills.

writing-kit-skills