skill-creator

To scaffold a new agent skill, this guided tool walks you through the entire creation process, helping you plan the structure, write the SKILL.md file, generate supporting files, and validate the final output against specification rules.

Best practices for writing agent skill instructions include using imperative voice, following strict naming conventions, structuring metadata correctly in frontmatter, and keeping descriptions concise to ensure proper agent comprehension and execution.

During skill creation, use the references directory to store domain knowledge and contextual data, while the scripts directory should hold executable automation code that the agent runs to perform specific tasks.

Yes, pyyaml is required as a dependency to create an agent skill, as it is necessary for parsing and validating the frontmatter metadata structure within the SKILL.md file.

Yes, you can validate your SKILL.md file against the Agent Skills specification by using the built-in validation step, which checks for proper frontmatter requirements, naming rules, and adherence to progressive disclosure principles.

Progressive disclosure in agent development is a design principle that structures skill instructions and supporting files to reveal information progressively, ensuring the agent receives only the necessary context at each stage of execution.