- Home
- Skills
- Productivity
- Idea Refine: Sharpen Raw Ideas
More screenshots
Works with the AI Agents you already use
Idea Refine: Sharpen Raw Ideas
Turn a vague idea into a decision-ready one-pager: reframe the problem, score the real directions, name the assumption that could kill it, and write the Not Doing list.
$5
Idea Refine: Sharpen Raw Ideas
Example session with this skill installed
I want to build an AI app that plans your meals for the week. Refine the idea before I start building.
- Read your context and instructions
- Compiled the idea refine
ONE-PAGER — produced by applying the skill.
Problem: How might we help two-income households who cook 4–5 nights a week decide what is for dinner in under five minutes a day, without a 40-question onboarding?
Direction: not a meal planner. Three directions were scored: full auto-planner (value 1 / feasibility 1 / differentiation 0 = 2 — cut; established apps own it and churn by week two), a pantry-first planner that builds the week from what is already in the fridge (2/2/2 = 6 — chosen), and a couples' swipe-to-agree picker (2/2/1 = 5 — parked as a later candidate).
Assumptions to validate: (1) decision fatigue, not recipe discovery, is the real pain — test with six household interviews narrating last Tuesday's dinner decision; (2) households will photograph fridge shelves weekly — test with a 10-user diary study and count week-two compliance; (3) "use what is open first" cuts grocery spend enough to notice — test with three weeks of receipts. Kill-fact: if households will not keep the pantry list current, this degrades into the generic planner it was meant to beat.
MVP scope: photograph the fridge, get a four-dinner plan built around what is open, one-tap swap per meal, shopping list for the gaps only. Not doing: calorie and macro tracking (a different product, with regulatory drag), auto-ordering groceries (partnership cost before the core loop is proven), breakfast and lunch (dinner is where the pain and the money are), recipe creation (link out — the bet is planning, not content).
Connects securely to your tools. The creator never sees your data.
About this skill
Most ideas die of vagueness, not badness: nobody can name the user, say what success looks like, or state what would prove the idea wrong. This skill runs a disciplined three-pass refinery that ends in a decision, not a brainstorm hangover. Pass one restates the idea as a "How might we…?" problem and forces five sharpening questions — specific user, observable success, real constraints, prior attempts, and why now — with a stop rule: if the user cannot be named, the deliverable is that finding, not polished fiction. Pass two opens the idea up with seven reframing lenses (invert it, remove a constraint, shift the audience, scale it 10×, and more), producing 5–8 variations that each carry their reasoning. Pass three converges: variations cluster into 2–3 genuinely distinct directions, scored on user value, feasibility, and differentiation, and the survivor faces the step most ideation skips — surfacing hidden assumptions, including the single kill-fact that would end the idea, each paired with an affordable validation test. The output is a one-page artifact with the chosen direction, an MVP scoped to test the core assumption (not a shrunken full product), open questions, and the most valuable section of all: an explicit Not Doing list with reasons, so trade-offs stop resurfacing as scope creep. You get the fill-in one-pager template, a refinement checklist, a readiness rubric, a full worked example (a plant-care app whose real idea turned out to be photo diagnosis, not reminders), and edge-case handling for solutions looking for problems and stakeholders' pet ideas.
- It cannot supply the missing user: if nobody can name a specific person with the pain, it stops and points at discovery work (interviews, observation) rather than refining around a placeholder audience.
- Its scoring is structured judgment, not measurement — it makes reasoning explicit and comparable, but the evidence behind a score is only as good as what you bring or go and get.
- It deliberately produces one page. Teams that need full requirements, market sizing, or a PRD should treat the one-pager as the input to that work, not a substitute for it.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 24 days ago
Permissions
No special permissions declared or detected