Runs the same loop as Anthropic's built-in skill-creator, capture intent, draft, test with and without the skill, review, optimize the triggering description, package, but never starts from a blank, generic placeholder. Before it asks what the new skill should do, it checks whether this install already carries a company profile; if not, it pulls from sibling deliverables, the 2nd Brain and Knowledge Graph vault, a persona.md, or interviews you directly for the vocabulary, systems, and real scenarios a plain skill creator never asks about. That context shapes the new skill's name, its trigger-forward description, its worked examples, and its test cases, so the finished skill reads like it was built for this business rather than adapted to it. Ships the full evaluation loop: paired with-skill and without-skill runs, a browser results viewer, quantitative benchmarking across iterations, description-triggering optimization against held-out queries, and a packaged .skill file ready to install.
- Ground: a client edition's company files, the knowledge graph vault, a persona.md, or a short direct interview, whichever answers first.
- Capture and draft: what the skill should do and when, written in the company's own vocabulary, then the full SKILL.md with examples from a real week.
- Test: paired with-skill and without-skill runs on realistic prompts, reviewed together in the results viewer, refined until the feedback comes back clean.
- Optimize and package: an optional description-triggering pass against held-out queries, then a packaged .skill file ready to install.
You provide
- What the new skill should do, and when it should trigger
- Wherever your company context already lives (a vault, a persona.md, a client edition) or a short interview if none exists
- Your review of the test-case outputs and, if you want it, the description-optimization eval set
You receive
- A drafted SKILL.md grounded in your company's vocabulary, systems, and real scenarios throughout
- evals/evals.json, a workspace of test runs, and a benchmark comparing with-skill against without-skill
- A packaged .skill file, plus a triggering-optimized description if you ran that pass
Download this skill and its install guide: https://exec-skills.paulcheek.com/skills/cheek-custom-skill-creator
How to use this skill
Once installed, start a new conversation and paste this prompt. The slash command invokes the skill directly; everything after it is your opening brief, so edit it to fit.
Sample output
A completed run of this skill for a sample company, Mastercard, exactly as the deliverable would arrive. Open it in a new tab.
The samples are illustrative demonstrations of the output format, built from public information. Mastercard is not affiliated with and has not endorsed these materials.