Skills

Two agent skills for building products: a blueprint to start, an audit to ship.

Version 1.1.0

What it is

Plain-Markdown skills on the open Agent Skills specification, so they run in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and the rest.

blueprint establishes a project's foundation with its owner in seven documents: vision, market, business, product, brand, stack and plan. production is an autonomous readiness audit: 39 parts, 753 evidence-backed checks, and a go or no-go for every surface, including MCPs, agents, self-hosted deployments, offline sync, real-time systems, devices and software supply chains.

The ventures on this site were blueprinted with the first, and BrainLLM went through the second before its launch.

Two skills from product foundation to release, with 39 audit parts and 753 checksInstall complete skill folders from the source repositoryIllustrative seven-document blueprint index with owner reviewSynthetic MCP audit: HOLD despite a verified score of 100, with 79 percent coverageComposed MCPRush publisher directory with blueprint, production and BrainLLM

What it’s for

For founders and coding agents establishing product intent before building, then assessing production readiness with evidence tied to the software surfaces being shipped.

  • Before you build

    Write down with the owner why the product exists, who it is for, its money and its plan, in seven documents.

  • Before you ship

    Audit the real software surfaces and get a go or no-go backed by evidence.

  • In whatever agent you use

    Both work in any agent that reads the open Agent Skills format.

Version
1.1.0
Licence
MIT
Install
npx skills add miisodev/skills · or the Claude Code plugin marketplace
Needs
An agent that reads and writes files · Python 3.8+ for the audit's scoring script, optional

How it’s made

  1. Intent, not design

    A blueprint sets why, for whom, limits, money and direction, and leaves design and implementation to the model that builds.

  2. Evidence or it isn't checked

    Every audit check cites what it found. The result is a JSON payload and a report ranking parts and surfaces by risk.

  3. No assumed layout

    Both skills find a project's intent wherever it is kept, rather than expecting a folder or a file name.

  4. Checked as a package

    A validator resolves all 753 checks, and behavioural tests run against both shipped examples before each release.

Disclaimer

  • An audit result applies to its stated scope and evidence. It does not replace specialist certification, and using a skill alone does not certify a product.
  • The MCPRush download currently delivers an older 1.0.1 bundle. Install from GitHub or skills.sh to get 1.1.0.

Support

Write to miisodev@gmail.com about using either skill, or an audit result you want explained.