skill-creator

Skill creation involves drafting a description, running automated eval loops to test trigger accuracy, and iterating through feedback to converge on a robust workflow. This structured process ensures skills are measurable, repeatable, and effectively capture user intents.

Packaging skills as portable artifacts requires consistently structured frontmatter, a SKILL.md body, and optional bundles like scripts, references, and assets. Validating these components ensures the skill is ready for repeatable deployment and sharing across teams.

Yes, pyyaml is required to parse and validate the skill frontmatter during the scaffolding and iteration process. It enables the automated evaluation loops that test trigger accuracy and stability across varied prompts.

Prompt engineering improves trigger accuracy by applying evaluation loops that test the skill description against varied prompts. Iterating through feedback and benchmarks refines the description, ensuring the skill activates reliably and produces useful outputs.

Yes, you can refine existing skills by applying the interview, implement, evaluate, and iterate process. Running automated eval loops on the current description identifies trigger accuracy gaps and tracks improvements through targeted feedback.

Bundle scripts, references, and assets when the skill workflow requires external logic or supporting data to process user intents. These optional components are validated alongside the frontmatter and SKILL.md to ensure consistent, repeatable deployment.