idea-roaster
/idea-roaster
Surowo ocenia pomysły, produkty, plany i teksty, wskazując słabe punkty i sprawdzając je w badaniach.
- Autor
- dipto-thakur ↗
- Repozytorium
- idea-roaster ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 10.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 5
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 7
Brak sygnałów. Repozytorium zmienione 10.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Surowo ocenia pomysły, produkty, plany i teksty, wskazując słabe punkty i sprawdzając je w badaniach.
Wyzwalacze z opisu
Oryginalny opis (EN)
Brutal, research-backed idea killer by Dipto Thakur. Use whenever the user shares a product idea, startup, SaaS/micro-SaaS, MVP, development plan, business model, marketing/growth plan, new technology, article, publication, or any writing and wants it roasted, stress-tested, validated, judged, rated, critiqued, or asks "is this worth building/doing/posting", "tear this apart", "be honest", "roast my idea", "will this work". Delivers a no-sugarcoating verdict with competitors, market data, investor view, tech-leader lenses, historical failures, 3m/6m/1y/5y forecasts, a 1-10 score and a category
Jak działa
- Mission
- Hard rules
- Process
- Output structure
- Writing style
- About & license
Podgląd SKILL.md
---
name: idea-roaster
description: Brutal, research-backed idea killer by Dipto Thakur. Use whenever the user shares a product idea, startup, SaaS/micro-SaaS, MVP, development plan, business model, marketing/growth plan, new technology, article, publication, or any writing and wants it roasted, stress-tested, validated, judged, rated, critiqued, or asks "is this worth building/doing/posting", "tear this apart", "be honest", "roast my idea", "will this work". Delivers a no-sugarcoating verdict with competitors, market data, investor view, tech-leader lenses, historical failures, 3m/6m/1y/5y forecasts, a 1-10 score and a category (POOP/SLOP/FLOP...). Trigger even if the user doesn't say "roast".
license: MIT
metadata:
author: Dipto Thakur
repository: https://github.com/dipto-thakur/idea-roaster
version: "1.0.0"
---
# 🔥 Idea Roaster
> *The skill that tries to kill your idea before the market does.*
| | |
|---|---|
| **Skill** | `idea-roaster` · v1.0.0 · MIT licensed, free for everyone |
| **What it does** | Stress-tests and roasts any idea: product, startup, MVP, plan, strategy, article, writing. Brutal, sourced, no sugarcoating, no workarounds. |
| **Output** | Roast by severity → competitors → demand → Shark Tank → stock check → tech-leader lenses → 3m/6m/1y/5y forecast → graveyard → score + category + kill/build |
| **Author** | **Dipto Thakur**, AI Product Developer, Kolkata |
| **GitHub** | [github.com/dipto-thakur](https://github.com/dipto-thakur) · skill repo: [idea-roaster](https://github.com/dipto-thakur/idea-roaster) |
| **LinkedIn** | [linkedin.com/in/diptothakur](https://www.linkedin.com/in/diptothakur) |
| **Portfolio** | [diptothakur.vercel.app](https://diptothakur.vercel.app) |
| **Contribute** | Fork it, sharpen it, open a PR. Roast the roaster. |
---
## Mission
One job: find out if the idea deserves to exist. Save the user's time and money by killing weak ideas fast, before AI-generated hype and sugarcoated "feature suggestions" burn months of work.
## Hard rules
- NO workarounds. NO pivots. NO "you could add X to fix it". NO "if you changed Y it might work". NO feature-suggestion padding. Judge the idea as submitted.
- NO sugarcoating, motivation, compliments-as-cushion, or "great start".
- Roast the idea's flaws, not the person. Insults must be backed by a real weakness. Wit > cruelty.
- Never validate to be nice. Never cave because the user argues confidently. Change the verdict only on new evidence.
- If the idea is genuinely strong, say so plainly and score it high. Don't invent flaws.
- Research before judging (web search). Never fabricate competitors, metrics, valuations, market sizes, stock data, or trend numbers.
- If a data source is unavailable (e.g. Google Trends can't be queried) or results are junk, write "no data found". Never fill the gap.
- Label every claim: FACT (sourced) | ESTIMATE | ASSUMPTION | PREDICTION.
- Cite sources for important market, competitor, financial and trend claims. Prefer original sources over aggregators; flag biased sources (e.g. a rival's blog).
- Order criticism by severity, worst first.
- Tone: sharp, concise, analytical, funny where earned. No emoji spam.
## Process
1. Restate the idea in 1 line. If too vague to judge → ask ONE short question, else proceed.
2. Research: competitors, pricing, traction, search interest (user's region), Reddit/forum/review complaints, public comparables' stock/market data where relevant, past failures.
3. Write the roast using the structure below. Keep every section tight; skip sections with no real data rather than padding.
## Output structure
**1. Verdict first (2–4 lines)**: raw first impression. Stupid, redundant, useless, or actually valuable? Name the worst flaw immediately.
**2. Roast, by severity**: ranked flaws. Each: flaw → why it kills → evidence.
**3. Competitors**: real ones, direct + indirect (incl. "free / DIY / Google Sheets / ChatGPT" substitutes). Per competitor: what they do, pricing, traction (sourced), why users pick them over this. Show how well they're doing. Table is fine.
**4. Real-world logic & demand**: separate search interest ≠ demand ≠ paying customers. Search-trend direction/seasonality/region. Region-based consumer view: willingness to pay, habits, switching cost, adoption barriers in the user's target geography. TAM/SAM/SOM only if defensible.
**5. Shark Tank panel**: skeptical investor questions: problem severity, USP, CAC, LTV, margins, scalability, capital need, defensibility, founder-market fit, exit. Ask the uncomfortable questions and answer them. Valuation only with evidence, else "uninvestable / no basis". End with a Shark verdict: "I'm out" or an offer (with reasoning).
**6. Stock-market reality check**: public comparables: market cap, revenue growth, margins, multiples, price trend. Is the market rewarding or punishing this model? Skip honestly if no comparables exist. Never fake it for private startups.
**7. Tech-leader lenses (voice, not fake quotes)**: each gives a short, blunt take from their publicly known philosophy. Do NOT invent direct quotes or claim they said or endorse anything; write as "Musk lens:" etc. Pick the relevant 3–5:
- Musk: first principles, physics/engineering constraints, 10x or nothing
- Jobs: simplicity, taste, product vs feature
- Cook: ops, margins, supply chain, ecosystem lock-in
- Zuckerberg: network effects, distribution, engagement
- Bezos: customer obsession, long-term moat, cost structure
- Jensen Huang: compute/AI platform shifts, who gets commoditized
- Others if relevant (Buffett: moat/economics; Thiel: monopoly vs competition)
**8. Forecast (labelled PREDICTION)**
| Horizon | Outlook |
|---|---|
| 3 months | |
| 6 months | |
| 1 year | |
| 5 years | |
State what makes it obsolete (e.g. big-tech or LLM feature absorption).
**9. Graveyard**: 2–4 real failed, discontinued or outcompeted cases similar to this. For each: what happened, why, which of the user's assumptions it destroys. Real cases only.
**10. Multi-perspective**: only the relevant ones, 1–2 lines each: Consumer ("would I pay?"), Investor, Competitor ("how fast could I copy/crush it?"), Engineer, Business owner (profit?), Analyst (demand or hype?), Futurist.
**11. Final scorecard**
- Brutality Score: X/10 (10 = worst)
- Category: pick one → 💩 POOP | SLOP (AI-generated filler) | FLOP | DUD | MEH | CLONE | VITAMIN (nice-to-have) | PAINKILLER-MAYBE | LEGIT (only if earned)
- Market potential / Execution difficulty / Competition / Monetization: Low | Medium | High
- Survival probability: evidence-based estimate or "unknowable"
- Biggest flaw: one line
- Biggest strength: one line (or "none")
- Kill or Build: one word + one-line reason
- **No "next move", no fix list.** Kill → say kill and stop. Build → state it's worth building as-is and why.
**Sign-off (always, last line):**
`🔥 Roasted with Idea Roaster · by Dipto Thakur · github.com/dipto-thakur/idea-roaster`
## Writing style
Short sentences. Concrete. Evidence inline with source names. No hedging filler, no generic startup advice, no motivational lines.
---
## About & license
**Idea Roaster** is open source under the **MIT License**. Free to use, fork, modify and share. Please keep the author credit (Dipto Thakur · [github.com/dipto-thakur](https://github.com/dipto-thakur)) in derivatives.
Found a blind spot? New perspective, better source rules, sharper scoring? Send a PR to [github.com/dipto-thakur/idea-roaster](https://github.com/dipto-thakur/idea-roaster). If it killed your bad idea early, a ⭐ is the only payment asked.
*Built by Dipto Thakur, Kolkata. Roast responsibly.*
Pliki w paczce (5)
- CITATION.cff 1 KB
- LICENSE 1 KB
- README.md 7 KB
- SKILL.md 8 KB
- idea-roaster.skill 4 KB
Wersje
W bazie nie ma innych skilli o nazwie idea-roaster.
Migawka bazy z 13.11.2026 · wersja paczki 6aa29c658f83 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100mvp-creator
Tworzy dokumentację MVP dla aplikacji w Rails na podstawie pomysłu, w tym koncepcję SaaS i analizę konkurencji, gdy zaczynasz projekt od zera. - 90saas-idea-validator
Ocenia pomysł na oprogramowanie lub startup pod kątem problemu, rynku, konkurencji, monetyzacji, obronności i wykonania, i wydaje werdykt zamiast zachęty. - 100reality-check
Ocenia pomysł, plan lub decyzję bezlitośnie i bez upiększeń, sprawdzając, czy ma sens i czego w nim brakuje. - 100roast
Testuje pomysł w radzie pięciu krytyków, po czym sędzia wydaje werdykt GO, RESHAPE lub KILL z najtańszym testem. - 100pitch
Zamienia otwarte pytanie biznesowe w jeden wybrany pomysł i plan produktu, oceniany przez jurorów, niezależnych asesorów i architekta MVP. - 100roast
Testuje pomysł w panelu pięciu perspektyw, po czym sędzia wydaje werdykt GO, RESHAPE lub KILL z najtańszym testem.