Do I need to buy anything to use the kit?

No. The kit, the method, and the skill catalog are open and MIT licensed. You need a Claude account to use the agent path, or just the document to work by hand.

What does my agent actually do with the file?

It follows the kit's agent protocol: interviews you about your processes, scores and ranks them, drafts your first five workflow files using the kit's template, and emits a manifest. It binds skills only after verifying they exist in the catalog, and it does not execute or publish anything; your team validates the workflows by hand first.

How is this different from the engines?

The kit produces your workflow library; the engines operate one. Same doctrine on both sides: held drafts, human merges, honest statuses. The kit is the do-it-yourself path, and the engines are the operated path for teams that want cadence and verification records without running it by hand.

Starter kit

Start here: standardize your operation with your own Claude

A free starter kit that turns the processes your team runs from memory into a written, validated workflow library. Point your Claude agent at one file and it interviews you, ranks your processes, and drafts your first five workflows. No signup, no engines required.

The one instruction

Paste this into a Claude conversation:

Fetch https://raw.githubusercontent.com/rampstackco/claude-skills/main/STARTER-KIT.md
and follow the agent protocol in section 6 to build a workflow library for my business.

Your agent will ask about your business, your recurring processes, and what has gone wrong recently. Twenty to forty minutes of answering questions produces your process inventory, a ranked scoresheet, five workflow files, and a manifest that tracks their status honestly.

How it works

You and your team
Your Claude

You and your team:Paste the one instructionFrom the start page
Your Claude:Interview, one question at a timeFetches the kit first
Your Claude:Score and rank your processesYou adjust the ranking
Your Claude:Draft five workflow filesAll status: template
You and your team:Cold run by a non-authorFlips to validated
You and your team:OperateAI drafts, humans publish
Operated layer, the enginesOptional, when validated

Your agent does the drafting. Your team does the validating. Nothing claims a status a cold run has not proven.

What the kit is

The kit is one markdown file in the open catalog, readable by you and executable by your agent. It carries the method RampStack uses on its own properties: inventory and score your processes, write the top five as stranger-runnable workflow files, validate each with a cold run by someone other than its author, and only then layer in AI assistance, phase by phase, with a human releasing everything that ships. Prefer to work by hand? The same file reads as a guide; the agent path is a convenience, not a requirement.

The three rules

Stranger-runnable or it is not done. If a step needs tribal knowledge, the knowledge goes in the document.
Statuses are honest. Every workflow is template, validated, or hardened, and nothing skips the cold run.
AI drafts, humans publish. Agent output lands as a held draft; a named person reviews and releases it.

Where the engines fit

The method is free and stays free, along with the full skill catalog it binds. The engines are the operated layer for teams that want the library run continuously: Krine decides, Tholo builds, Basano proves, and every change is a held draft a human merges. When your library is validated and you want it operated, read the reference architecture at /engines/deployment.

View the kit on GitHub Browse the skill catalog See the operated layer

FAQ

Do I need to buy anything to use the kit?
No. The kit, the method, and the skill catalog are open and MIT licensed. You need a Claude account to use the agent path, or just the document to work by hand.
What does my agent actually do with the file?
It follows the kit's agent protocol: interviews you about your processes, scores and ranks them, drafts your first five workflow files using the kit's template, and emits a manifest. It binds skills only after verifying they exist in the catalog, and it does not execute or publish anything; your team validates the workflows by hand first.
How is this different from the engines?
The kit produces your workflow library; the engines operate one. Same doctrine on both sides: held drafts, human merges, honest statuses. The kit is the do-it-yourself path, and the engines are the operated path for teams that want cadence and verification records without running it by hand.