Blueprints Skill — AI skill for Claude Code

Blueprints — Cross-Tool Product Development Skill

[](https://skills.sh/siddheshbmane/blueprints-skill)

**A portable Agent Skill that turns any AI coding tool into a documentation-first, qualify-first SaaS product builder.**

Built from a real-world SaaS playbook (~8,700 lines) and distilled into a lean `SKILL.md` + deep references that work in **OpenCode, Claude Code, Codex, Grok, Gemini CLI, Cursor**, and any agent that supports the [Agent Skills](https://agentskills.io) `SKILL.md` standard.

It enforces Siddhesh's blueprint: documentation before code, intake questionnaires, multi-tenant design from day 1, no-AI-slop UI, staging-before- production deploys, browser-based user-flow E2E tests, and a mandatory task report after every change.

---

What it does

When you load this skill, your agent follows a proven product-development system:

1. Ask the Project Intake questionnaire (12 topics) + Design Intake questionnaire (8 questions).
2. Generate Phase 0 docs — SCOPE, SCHEMA, API, WIREFRAMES, ADR, BRAND (no code until you approve).
3. Build a modular monolith → monorepo → hardening → mobile → GTM.
4. Deploy to staging first, verify the live URL, run real-user E2E tests.
5. Gate production behind an explicit "Ask User" gate — never auto-merges to main.
6. Report after every task (changes, deploy, tests, brand check, known issues).

Plus an optional **product due-diligence protocol** (`references/dd.md`) to evaluate competitors, acquisitions, or your own roadmap bets.

The 15 Golden Rules

1. Read heart.md (the full blueprint) FIRST, entirely, before anything.
2. Ask the PROJECT INTAKE questionnaire before any work.
3. Ask the DESIGN INTAKE questionnaire before any UI work.
4. Generate ALL Phase 0 docs. Get user approval before coding.
5. Multi-tenant from Day 1 — organizationId on EVERY table.
6. No AI slop design. No purple gradients. Award-winning UI only.
7. Ser