skill-creator

To turn ad-hoc task patterns into reusable agent skills, use this tool to capture intent, plan with files, implement resources, validate frontmatter, and package the workflow into a zip artifact. It standardizes scattered workflows into modular Skill Units for process automation and document processing.

The standard structure to package an agent skill requires a SKILL.md file with required frontmatter for discovery and triggering. It also supports optional resource directories including scripts, references, and assets to organize tooling and outputs for distribution.

To create a new skill and initialize directories, run init_skill.py with the desired skill name and path. This generates the standardized skeleton where you can tailor SKILL.md and add any needed resources before validation and packaging.

Skill creation and frontmatter validation require the yaml dependency to parse configuration correctly. The workflow operates within the agent ecosystem, using a plan-with-files approach to enforce required metadata and optional resource directory structures before packaging.

The best way to organize scripts and references when designing agent skills is using the optional resource directories. Place tooling in scripts, reference materials in references, and outputs in assets to maintain a standardized structure that enables reliable discovery and triggering.

You should not use a standardized skill template for process automation when your workflow cannot be modularized into discrete intent, planning, and validation steps. Complex ad-hoc workflows lacking defined triggers or reusable patterns are unsuitable for this packaging approach.