BuilderOS - Setup Guide¶
Source: buildgreatproducts/builder-os via skills.sh - 15.7K combined installs across 10 indexed listings; first seen Oct 9, 2026 (extended-query probe)
GitHub: buildgreatproducts/builder-os (229 stars, MIT license; last pushed Jul 7, 2026; formerly PLAID - the PLAID skills were refactored into standalone BuilderOS skills; layout skills/<name>/SKILL.md)
Category: Product Development / Builder Workflow
Quality Tier: 🟡 Beta - 229-star MIT suite; formerly PLAID; last pushed Jul 2026; all sampled verdicts Pass
BuilderOS, from publisher BuildGreatProducts, is the successor to PLAID (Product Led AI Development): a suite of ten agent skills that give product builders a repeatable system for ideating, designing, and building software with AI. The premise is discipline - AI coding agents are fast, and speed without process produces features that compile rather than features that work. BuilderOS encodes product-team habits (plan-driven work, mandatory review, end-to-end testing, honest reporting) as skills an agent follows automatically, turning a coding agent from autocomplete into a disciplined collaborator.
The skills chain through documents. Each one is fully standalone, but each writes its output to a shared docs/ folder at the project root, and downstream skills pick those documents up automatically: idea-generator and idea-validator feed product-planner, product-planner feeds the build skills, and design-system feeds the craft and build layers. No single skill dominates the install mix; all ten listings sit above 1,000 installs, which suggests the suite is adopted across the whole lifecycle.
Installation¶
BuilderOS installs through the skills CLI from your project root:
# Everything in one go
npx skills add BuildGreatProducts/builder-os
Individual skills are self-contained; install just the one you need:
npx skills add BuildGreatProducts/builder-os/skills/idea-generator
npx skills add BuildGreatProducts/builder-os/skills/product-planner
npx skills add BuildGreatProducts/builder-os/skills/build-loop-claude-code
npx skills add BuildGreatProducts/builder-os/skills/design-system
Or select by skill name:
npx skills add BuildGreatProducts/builder-os --skill product-planner
Manual alternative: clone the repo and copy the skill folders you want into your project's skills directory (for example .claude/skills/ for Claude Code). You only need the build loop for the coding agent you actually use - Claude Code, Codex, or Cursor.
What It Provides¶
Install counts from the Oct 10, 2026 skills.sh snapshot:
| Skill | Installs | Use For |
|---|---|---|
| design-system | 1,820 | Turn screenshots, mockups, or Figma URLs into a design system: tokens plus rationale in design.md, mirrored by a live design.html style guide |
| product-planner | 1,684 | Structured vision intake (8 sections) that generates strategy, a coding-agent-ready spec, and a phased roadmap with task checkboxes |
| idea-validator | 1,674 | Pressure-test an idea: core assumption, ranked fatal flaws, real competition, first 10 customers, a 2-week MVP test, and a blunt verdict |
| launch-checklist | 1,672 | Audit the codebase (stack, services, env vars, payments, deploy config) and write a step-by-step path from "works on my machine" to "customers can use it" |
| idea-generator | 1,671 | Guided discovery of a product idea: mine what you already know, synthesize 3-5 candidate directions, score them, and sharpen the winner |
| build-loop-claude-code | 1,642 | Review-gated build loop for Claude Code: every increment is built, reviewed, tested end to end, then fixed before anything ships |
| build-mvp | 1,610 | Execute the entire roadmap end to end, testing and verifying each task, then finish with an initial git commit |
| build-loop-codex | 1,482 | The same build-review-test loop adapted to OpenAI Codex CLI's review tooling |
| build-loop-cursor | 1,439 | The same loop adapted to Cursor's review tooling |
| design-better | 1,022 | Craft layer for UI: 50 numbered UX/UI heuristics plus a pre-flight checklist, from hierarchy and typography to motion and WCAG 2.2 AA |
No indexed listings fall below the threshold.
Why This Matters for Hermes Agents¶
Coding agents are fast, but speed without discipline produces features that compile rather than features that work - BuilderOS's own framing of the problem. The suite targets that gap with encoded process rather than more prompting: plan-driven work, mandatory review, end-to-end testing, and honest reporting become skills the agent follows automatically. The document chain is what makes multi-step projects tractable: each skill reads upstream documents from the project's docs/ folder, so context carries across sessions and across agents instead of living in one chat window. Three build-loop flavors cover Claude Code, Codex, and Cursor, and build-mvp is agent-agnostic, so the same lifecycle works whichever coding agent runs underneath. The craft layer matters for agent output quality: design-better's heuristics and design-system's tokens give review passes something concrete to check against. Everything is MIT licensed and installs as plain markdown skills through the standard CLI.
Usage¶
| You say | What happens |
|---|---|
| "Help me find a product idea" | idea-generator mines your context, synthesizes 3-5 directions, scores them on a five-axis scorecard, and writes docs/product-idea.md |
| "Validate my idea before I build" | idea-validator ranks fatal flaws, maps competition (including doing nothing), plans the first 10 customers, and delivers a strong / weak / pivot verdict |
| "Plan my product" | product-planner runs the 8-section vision intake and generates VISION.md, product-vision.md, prd.md, and product-roadmap.md |
| "Create a design system from this screenshot" | design-system produces design.md tokens plus a live design.html style guide the whole suite can implement from |
| "Build my MVP" or "run the build loop" | build-mvp executes the whole roadmap task by task; the build loop instead ships tighter review-gated increments - built, reviewed, tested, then fixed |
| "Make this UI feel more designed" | design-better applies 50 UX/UI heuristics and a pre-flight checklist, flagging missing tokens instead of inventing values |
| "Create my launch checklist" | launch-checklist audits the actual codebase and writes a step-by-step launch path with time estimates, costs, and success checks |
Verification¶
# Confirm the skills landed (Claude Code example; adjust to your agent's directory)
ls ~/.claude/skills/ | grep -i -E "idea-validator|product-planner|build-loop"
# Review a skill straight from GitHub before installing (layout: skills/<name>/SKILL.md)
curl -s https://raw.githubusercontent.com/buildgreatproducts/builder-os/main/skills/design-system/SKILL.md | head -20
Security¶
skills.sh verdicts for sampled skills (verified Oct 10, 2026) - re-check the security pages on skills.sh before production use:
| Skill | Gen Agent Trust Hub | Socket | Snyk |
|---|---|---|---|
| design-system | Pass | Pass | Pass |
| product-planner | Pass | Pass | Pass |
| design-review | - | - | - |
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.
Limitations¶
- Last pushed July 2026: recent enough to work as designed, but the suite is not under active daily development.
- Agent-specific build loops: install the one matching your coding agent (Claude Code, Codex, or Cursor); build-mvp is the agent-agnostic route.
- Security sampling is partial: design-review returned no verdicts from any engine in the sampled set.
- The lifecycle assumes a docs/ folder workflow - run the skills from the project root so the document chain works.
- Skills are standalone, but the payoff compounds only when upstream documents exist; running the build skills without a roadmap skips the pre-fill behavior.
- Snapshot data, verified Oct 10, 2026: 15,716 combined installs across 10 indexed listings; 229 GitHub stars; MIT; last pushed Jul 7, 2026. Counts drift over time.
Related¶
- PM Skills - Product Management Lifecycle Suite Setup - product-management counterpart across the same lifecycle stages
- Compound Engineering Plugin - Every Inc Workflow Suite Setup - workflow-suite sibling for engineering teams
- ClawFu Skills - 175 Marketing Methodologies for AI Agents Setup - marketing layer for the launch phase this suite ends on
- Skills Catalog | Skills Marketplace
Pro Tips¶
- Start anywhere in the lifecycle: the README maps entry points - "help me find an idea", "validate my idea", "plan my product", "create a design system from this image", "build my MVP", "run the build loop", "design better", "create my launch checklist".
- Let the docs/ chain work: run upstream skills first, or drop existing documents into docs/, so each skill pre-fills its questions.
- Install only the build loop for the coding agent you actually use; add build-mvp when you want the whole roadmap executed in one pass.
- Design System accepts images, mockups, and Figma URLs, and its design.md becomes the token source for Product Planner, Build MVP, Build Loop, and Design Better.
- Trigger the craft layer for UI work with "design better" or "make this feel more designed" - it flags missing tokens as New Patterns instead of inventing values.
← Skills Catalog | Skills Marketplace → Powered by CorpusIQ