AI Blueprint Assistant

What this skill does

• Detects the project type adaptively (software AND garden, event, marketing, DIY, course, business, club …) and activates fitting modules
• Five modes via a router: Direct plan · Quick intake (5 questions) · Pro intake · Research first · Improve existing project
• Marks assumptions visibly and prioritizes open questions instead of guessing blindly
• Separates MVP, expansion and vision; hard-caps version 0.1 and names non-goals (anti-overengineering)
• Delivers target picture, scope, structure, resources, roadmap, task list, risks and Definition of Done
• Blueprint Score (0–100) with strengths and concrete improvements
• Research mode: flags verify-first areas (tech, prices, law, local/safety factors) or uses a web tool
• Safety logic: friendly-brakes on risky/abusive projects and marks expert-review points
• Produces a directly usable AI handoff for ChatGPT, Claude, Cursor, Codex, Copilot or Gemini
• Generates project files on request (PROJECT_BLUEPRINT, PROJECT_MEMORY, DECISION_LOG, TASKS, AGENTS.md, CLAUDE.md, .cursor/rules, copilot-instructions, MATERIAL_LIST …)

Description

AI Blueprint Assistant is a skill for AI-ready project planning and scoping — it turns a rough idea (software or not) into a complete project blueprint with scope, structure, roadmap, risks, quality gates and a copy-paste AI handoff. It builds a project brain instead of just a prompt: the context a human and an AI need to carry a project forward reliably, safely and without chaos. Domain-universal via activatable modules: it detects the real project type (app, website, game, marketing, event, garden, DIY/renovation, course, book, business, organization, mixed forms) and combines the fitting modules instead of forcing a rigid list. Five working modes via a router: Direct plan (sensible, flagged assumptions), Quick intake (5 questions), Pro intake (full intake), Research first (current best practices, prices, local & legal factors — with a web tool or flagged verify-first) and Improve existing project (diagnose and rebuild a chaotic project). Every blueprint separates MVP, expansion and vision, hard-caps version 0.1, names non-goals and, on request, gives a Blueprint Score (0–100) with strengths and concrete improvements. Safety & honesty throughout: it friendly-brakes on harmful, deceptive, spam/scam, privacy-violating or physically dangerous projects, marks legally, financially, health- or structurally risky areas as verify-first and invents no numbers or sources. It closes with a directly usable AI handoff; on request it generates named project files such as PROJECT_BLUEPRINT.md, PROJECT_MEMORY.md, DECISION_LOG.md, RISK_REGISTER.md, TASKS.md, AGENTS.md, CLAUDE.md, GEMINI.md, .github/copilot-instructions.md, .cursor/rules/project.mdc or non-software files (MATERIAL_LIST.md, TIMELINE.md, CHECKLIST.md).

Examples

club-app/
  docs/  PROJECT_BLUEPRINT.md  DECISION_LOG.md  TASKS.md  TEST_PLAN.md
  src/   app/ components/ lib/ server/
  AGENTS.md  CLAUDE.md  README.md

What it does not do

• Doesn't treat every project as software and forces no rigid category list
• Invents no facts, prices, statistics or sources — flags research need instead
• Gives no final legal, tax, medical or financial advice; no definitive how-to for dangerous electrical/structural/gas/chemical work
• Won't help with harmful, deceptive, spam/scam or privacy-violating projects
• Won't drown you in questions or produce unstructured walls of text; won't overengineer the MVP

Compatibility & tech

Tested (internal)

5 scenarios

Recommended runtime

Strongest available model; for research mode one with a web/search tool.

Modes

Direct plan · Quick intake (5 questions) · Pro intake · Research first · Improve existing project

Inputs

Project idea (text) · Answers to follow-up questions · Existing project / file tree · Key data (budget, deadline, audience)

Output format

Structured blueprint (project type, assumptions, scope/non-goals, structure, roadmap, risks, DoD, Blueprint Score) + copy-paste AI handoff; on request named project files (PROJECT_BLUEPRINT, AGENTS.md, CLAUDE.md, .cursor/rules, MATERIAL_LIST …)

Subcategory

Project planning & AI handoff

License

Proprietär

Security profile

What you get

• ai-blueprint-assistant-1.1.0/8 files
  • .claude-plugin/marketplace.json
  • plugins/ai-blueprint-assistant/skills/ai-blueprint-assistant/8 files

    • SKILL.md
    • manifest.json
    • references/6 files

      • ai-blueprint-assistant-blueprint-template.md
      • ai-blueprint-assistant-checklist.md
      • ai-blueprint-assistant-domain-modules.md
      • ai-blueprint-assistant-handoff-exports.md
      • ai-blueprint-assistant-intake-questions.md
      • ai-blueprint-assistant-safety-checks.md
  • .agents/skills/ai-blueprint-assistant/→ universal — same content (Codex, Cursor, Copilot, Gemini, Windsurf, Cline)
  • LICENSE.txt

Installation

Also works as a chat prompt

Note

Plans and hands off — it doesn't build the project itself. Strongest with a few data points (goal, audience, budget/deadline); without them it works from clearly flagged assumptions. Research on current/local facts only with a web tool, otherwise marked verify-first.

Changelog

• v1.1.022.06.2026Platform-native output: now actively uses your platform's native formatting — tables, checklists, code blocks and clickable/button options, with a numbered fallback on non-interactive platforms. Plus internal metadata that prepares the upcoming agent builder.
• v1.0.022.06.2026First release: domain-universal AI project planner to the AUTHORING v2 standard — five modes (Direct/5 questions/Pro/Research/Improve existing), activatable domain modules, MVP/expansion/vision split, Blueprint Score, safety & research logic, AI handoff + file exports (AGENTS.md/CLAUDE.md/Cursor/Copilot/Gemini/Codex). 6 references (output template, domain modules, question bank, handoff exports, safety checks, copy checklist), realistic examples incl. a calibration case and internal evals.

Frequently asked questions