skill-creator

To create validated LLM skill packages, you need a structured approach that enforces consistency. This skill scaffolds self-contained directories with SKILL.md, scripts, and references, applying a 9-block methodology to ensure packages pass automated validation and human-executed swap tests before deployment.

Swap testing in LLM development is a validation mechanism where a generated skill package is executed to confirm it produces reliable, consistent results. It is needed when creating or modifying AI skills to ensure the output meets strict automated validation checks and performs as expected in production environments.

Scaffolding a new AI skill involves following a strict 9-block process covering domain scope, constant extraction, process definition, workflow design, organization, and rules. This framework generates self-contained skill directories with the necessary scripts and references to ensure structured, validated execution.

Python 3 is required to run the automated validation framework for skill creation. The validation process uses scripts like quick_validate.py to check the generated skill packages, ensuring the 9-block methodology is correctly applied and the package passes automated checks before delivery.

You can modify existing AI skills and maintain validation by applying the same 9-block methodology used for new creation. The framework enforces strict block-based rules and runs automated validation scripts to ensure any modifications to the skill package remain consistent and swap-testable.

The best way to structure prompt engineering workflows is using a block-based development framework that defines domain scope, extracts constants, and designs processes. This methodology ensures the generated skill packages are self-contained, validated, and produce reliable outputs during human-executed swap tests.