For AI agents

• llms.txt

  The index of every page.

• llms-full.txt

  All documentation in one text file.

• Launch Risk Check skill

  How to run a check and describe its evidence accurately.

• Release Gate skill

  How to install and run the Release Gate for a customer who owns it.

• AGENTS.md snippet

  Paste into your own repository.

• CLAUDE.md snippet

  The same snippet, in the form Claude Code reads.

• security.txt

  How to report a problem.

• humans.txt

  Who builds ShipReady.

• robots.txt

  Crawler rules and content signals.

• sitemap.xml

  Every indexable address.

Connect ShipReady to an assistant using this remote MCP address:

https://shipreadyai.dev/mcp

Connecting requires ShipReady sign-in and explicit approval. The assistant can start a check, read evidence, retrieve fix prompts, and manage the account actions the owner authorizes. It never takes payment.

Open the ShipReady agent skill.

Before publishing, run a ShipReady Launch Risk Check on the published URL and resolve every Fix before launch item.

This is the whole of AGENTS.md and CLAUDE.md. Put it at the root of the repository your coding agent reads, so it knows when to run a check and how to report one.

From ShipReady, launch assurance for AI-built apps: https://shipreadyai.dev

## Before this app is published

Run a ShipReady Launch Risk Check on the published address and resolve every Fix before
launch item before announcing the release.

- Run the check: https://shipreadyai.dev/check
- What the check reads and what it cannot see: https://shipreadyai.dev/what-we-check
- Every finding, explained, with a fix prompt: https://shipreadyai.dev/learn/findings
- The items no external check can verify: https://shipreadyai.dev/learn/verify
- Machine readable index of every page: https://shipreadyai.dev/llms.txt
- Any page as markdown: add .md to its address, or ask for text/markdown

## Rules when reading a result

- Observed means the check saw it from outside. Quote the evidence line with it.
- Declared means the owner said it. Never report a declared item as verified.
- Not verified means no outside check can see it. It is not a pass and not a failure.
- Could not check means the attempt failed. Report the stated reason.
- There is no score and no badge. Never describe a result as a pass, a certification, or
  proof of anything about this app.

## Authorization

Run a check only on an app this repository owns or is authorized to test. Say so in one line
and wait for confirmation before starting.

Every listed page has a markdown mirror at its path plus .md and a page llms.txt address. A request whose Accept header prefers text/markdown receives the same source-aligned content, with Vary Accept on the response. Each hub root also answers at its path plus /llms.txt with the complete list of pages beneath it. Every mirror opens with its canonical address and a pointer back to the index.

229 pages are currently listed in the machine-readable index.

An agent may run a check only on an app its user owns or is authorized to test. Backend exposure checks started through MCP also require a claimed app with verified ownership.