idea-os
April 18, 2026 · View on GitHub
An operating system for turning raw ideas into build-ready plans — as a Claude Code / Claude-compatible skill.
You drop in a rough problem statement or idea. It runs a five-phase pipeline (clarify → research → PRD → plan) and writes four files you can take straight to building:
questions.md— sharp, tier-adapted clarifying questionsresearch.md— market, competitors, SWOT, JTBD, distribution, trends, insightsPRD.md— problem, users, scope, non-goals, metricsplan.md— user journey (mermaid), platform/stack recommendations, phased build with kill criteria
Why this exists
Coding is having a moment. Non-developers are opening terminals. Experienced PMs are shipping side projects. The hard part isn't the code anymore — it's the gap between "I have an idea" and "I know exactly what to build, for whom, how, and how I'll know it's working."
Most PRD tools stop at the document. idea-os goes further — it ties research to PRD to phased plan, with opinionated platform/stack recommendations, a mermaid user-journey diagram, kill criteria per phase, and first-100-users distribution. It adapts depth to the idea's complexity and the builder's experience so a first-time builder isn't buried in jargon and a founder gets full rigor.
What you get
- Clarifying questions (4–18, complexity-adjusted) that actually change the answer
- Deep research — TAM/SAM/SOM (top-down + bottom-up), segmentation, competitor matrix with positioning map, SWOT, JTBD, distribution strategy, why-now, 3–7 non-obvious insights
- PRD — falsifiable problem statement, named personas, ranked JTBD, leading/lagging metrics, explicit non-goals
- Execution plan with:
- User journey (text walkthrough + mermaid diagram that renders on GitHub)
- Platform recommendation (web / mobile / desktop / CLI / extension / hybrid) tied to research findings
- Stack recommendation (conservative / modern / cutting-edge matrix)
- Phased build (MVP → v1 → target state) with kill criteria and per-phase distribution
- Concrete next steps for the next 1–2 weeks
Installation
Claude Code
Clone into your skills directory:
git clone https://github.com/Slashworks-biz/idea-os.git ~/.claude/skills/idea-os
The skill activates automatically when you describe an idea or problem that needs the full pipeline.
Claude.ai
Zip the repo contents and upload as a project-level skill, or paste SKILL.md into a project description.
Trigger phrases
- "I have an idea for…"
- "Help me build X"
- "Validate and plan this concept…"
- "What should I build?"
- Any problem statement where the user wants the full multi-phase artifact output (not a quick one-shot answer)
How it works
Five phases, in order:
- Triage — classify idea complexity (C1 simple / C2 moderate / C3 complex) and jargon level (plain / standard / full). Depth and vocabulary scale accordingly.
- Clarify — write
questions.md, wait for answers. Skip if the brief is already dense or in autonomous mode. - Research — write
research.mdusing real signal (WebSearch + WebFetch). No hallucinated stats; sources or[assumption]flags. - PRD — write
PRD.md. - Plan — write
plan.mdwith user journey, platform/stack picks, phased build, metrics, kill criteria, distribution, next steps.
See SKILL.md for the full spec.
Repo layout
idea-os/
├── SKILL.md # skill entry point (trigger + workflow)
├── references/
│ ├── clarifying-questions.md # tiered question bank by idea type
│ ├── research-frameworks.md # JTBD, SWOT, Porter 5, PESTEL, Blue Ocean, Wardley
│ ├── market-analysis.md # TAM/SAM/SOM, segmentation, WTP
│ ├── competitor-analysis.md # direct/indirect/substitute, positioning map
│ ├── distribution.md # first-100-users channel-fit patterns
│ ├── prd-structure.md # PRD anatomy and anti-patterns
│ ├── mvp-slicing.md # vertical slice, walking skeleton, concierge MVP, RICE/ICE/Kano
│ ├── metrics-framework.md # North Star, AARRR, HEART, counter-metrics
│ ├── platform-recommendation.md # web/mobile/desktop/CLI/ext decision framework
│ ├── stack-recommendation.md # conservative/modern/cutting-edge defaults
│ └── user-journey.md # text + mermaid guidance
├── assets/
│ ├── questions-template.md
│ ├── research-template.md
│ ├── PRD-template.md
│ └── plan-template.md
├── examples/
│ └── example-notion-for-dog-trainers.md # reference-only T2/S2 walkthrough
└── README.md
Philosophy
- Specifics beat generics. A SWOT bullet that could apply to any product in the category is dead weight.
- Non-goals are diamonds. The PRD lives or dies by what you refuse to build in v1.
- Kill criteria up front. Every phase has a "stop here" threshold set before hope distorts judgment.
- Evidence over confidence. Every factual claim is sourced or flagged
[assumption]. - Distribution deserves product-level rigor. Where the first 100 users come from is a product decision, not a marketing afterthought.
- Jargon only where earned. Plain-level readers never see "Kano model" without an explanation. Full-level readers get the full vocabulary.
- MVP is a vertical slice, not a feature list. One user, one job, end-to-end.
Example
See examples/example-notion-for-dog-trainers.md for a complete mid-complexity walkthrough — original brief → questions → research → PRD → plan. It's a reference-only file showing the quality bar; you don't need to read it unless you're calibrating the skill.
Contributing
Issues and PRs welcome. Especially useful:
- Additional domain-specific question banks (healthtech, fintech, climate, etc.)
- Example walkthroughs at different complexity levels
- Sharper competitor-analysis templates for specific categories
Related skills
idea-os sits alongside (does not replace):
idea-refine— quick feedback on a single idea; use this when you don't want four filesproduct-management— PRD editing and ongoing product work for an already-scoped productdesign-thinking— methodology, complementary toidea-os's artifact pipelinesenior-software-engineer— downstream ofidea-os; once you have the plan, this helps you build it
License
MIT — see LICENSE.
Credit
Built under Slashworks with the Claude Agent SDK.