ck:skill-creator

To create Claude skills optimized for benchmark accuracy, use this skill to generate packages with deterministic instruction structures, high-quality activation metadata, and numbered workflows that align with Skillmark benchmarks for reliable execution.

Progressive disclosure in Claude skill authoring is a workflow design where bundled scripts, references, and assets are loaded dynamically based on clear rules, preventing context bloat while ensuring Claude has access to necessary resources.

You can add a security policy to your Claude skill template by using the skill's built-in safety-focused templates, which require a mandatory scope declaration and include a comprehensive security policy section to enforce prompt safety.

Before generating a Claude skill package, you need a populated skill root directory containing a SKILL.md file with frontmatter specifying the skill name and description, along with optional bundled scripts, references, and assets.

Yes, you can update an existing Claude skill to improve benchmark accuracy by applying this skill's benchmark-oriented structure, which refines the instruction determinism and activation metadata of your current skill packages.

Claude skills may not activate reliably if they lack high-quality activation metadata or a deterministic instruction structure. This skill resolves the issue by applying benchmark-aligned accuracy standards and progressive disclosure workflows.