skill-creator

To structure a Codex skill, create a SKILL.md entry point with valid YAML frontmatter and write concise, token-efficient operational instructions. This ensures the skill is self-contained, discoverable, and reliable for agents to execute.

Organize bundled resources by placing executable code in scripts/, supporting documentation in references/, and static media in assets/. This progressive disclosure structure keeps SKILL.md concise while ensuring necessary components are readily available.

Maintain UI metadata consistency by generating or updating agents/openai.yaml when creating a skill. Use hyphen-case for skill naming and ensure the metadata accurately reflects the skill's operational instructions and intended agent behavior.

Yes, validation and safe forward-testing are required for new skill scaffolding to catch issues early. This approach verifies workflow integrity and ensures the skill operates as intended without leaking intended fixes into the testing environment.

Allow appropriate degrees of freedom by writing concise instructions that guide the workflow without over-constraining the agent. This balance ensures the skill remains flexible enough to handle variations while maintaining strict validation integrity.

A Codex skill fails to load properly when the SKILL.md entry point contains invalid YAML frontmatter or lacks token-efficient instruction writing. Ensure the skill folder structure aligns with required metadata formats and validation rules.