← All productsAgent Foundry

Skills Engineering Kit

A complete methodology for building Claude Code skills that actually fire, actually work, and stay small enough to trust.

What's included

Every file that ships inside the zip — this is the manifest, not a highlight reel.

Skill anatomy
anatomy/skill-anatomy.md

The complete shape of a SKILL.md file — every frontmatter field, body structure, and the references/ and scripts/ subdirectory pattern, with hard rules for what belongs where.

Skeleton templates
anatomy/templates/*

Three ready-to-adapt SKILL.md skeletons covering the shapes almost every real skill reduces to: a step procedure, a lookup reference, and a CLI/script wrapper.

Description engineering
triggers/description-engineering.md

The craft of the description field — trigger phrase coverage, when-to-use vs when-not-to, anti-patterns, and before/after rewrites. A perfect skill behind a bad description never runs.

Progressive disclosure
progressive-disclosure/progressive-disclosure.md

Keeping SKILL.md small and pushing depth into references/, with token-budget reasoning and a worked example splitting an 800-line skill into a lean body plus three reference files.

Testing methodology
testing/testing-skills.md

Trigger probes to confirm a skill fires, execution tests to confirm it works, a regression checklist for edits, and adversarial testing for hook-adjacent skills.

Lint harness
testing/skill-test-harness.sh

A dependency-free bash script that lints every SKILL.md under a directory — frontmatter, name/directory match, description bounds, stray TODOs, orphaned reference files.

Worked examples
worked-examples/*

Two complete, production-quality skills (changelog-writer, pr-description) built with this exact methodology, ready to use as-is or as a model.

Pre-ship checklist
CHECKLIST.md

A one-page checklist to run against any new skill before it ships.

Pricing

Same download for every tier below. The license grant scales with your team size.

Standard

One license, whole methodology

$29 one-time

One person. Unlimited personal and commercial projects.

FAQ

QWhat exactly is in the download?

AA zip containing everything in the table above: the anatomy guide, three skill skeletons, the description-engineering and progressive-disclosure guides, the testing methodology, a bash lint harness, two complete worked-example skills, and a pre-ship checklist. About 15 files of self-contained Markdown and bash.

QDo I need the Autonomous Agent Ops Kit too?

ANo — this kit is fully standalone. They complement each other (ops patterns vs skill-building methodology) but neither requires the other.

QDo I need Claude Code specifically?

AYes for the skill format itself — SKILL.md is a Claude Code convention. The methodology chapters (trigger design, progressive disclosure, testing) transfer to any agent framework with routable tool/procedure descriptions.

QCan I resell or repackage the kit?

ANo. You can use it in unlimited personal and commercial projects — including skills you ship to clients — but you may not redistribute or resell the kit itself. Full terms in the included LICENSE.md.

QWhat's the refund policy?

A14 days, no questions asked, since this is a digital product delivered instantly. See /refunds for details.