skill-creator

Creating a skill with YAML frontmatter and trigger phrases involves drafting an SKILL.md file that defines operational instructions, trigger coverage, and verification hooks. The draft is then audited against an evidence-based rubric and revised once to maximize compliance and score.

A rubric audit for skill authoring is an evidence-backed scoring process that evaluates a draft skill against explicit criteria, including trigger coverage, anti-undertrigger wording, size fit, rationale for non-obvious instructions, and verification hooks.

Writing operational instructions to prevent undertrigger behavior requires applying specific anti-undertrigger wording constraints within the skill draft. The rubric audit then checks the draft's trigger phrases and operational instructions against these evidence-backed criteria.

Skill authoring with TBM conventions accounts for ES5 and HTML surfaces by applying forge conventions and ES5 awareness during the skill scaffolding process. This ensures the generated SKILL.md aligns with UI considerations and repository constraints.

The best way to verify a skill meets repository conventions is to audit the drafted SKILL.md against a strict rubric with evidence requirements. This process checks TBM-convention alignment, trigger phrases, and verification guidance.

You should use an automated revision loop for skill scaffolding when your draft fails rubric criteria and needs a single-pass revision to address gaps. The loop stops early if score gains diminish, emitting the final artifact with a scorecard.