Hunt Opportunity

---
name: hunt-opportunity
description: Web recon for business opportunities — find real, cited painpoints and money already in motion instead of inventing ideas from imagination. Use when FOUNDER mode starts, when a venture is pivoting, or when the user asks "what business should I start", "find me an idea", "หาไอเดียทำเงิน", "what problem should we solve". Runs WebSearch/WebFetch across complaint-dense and money-adjacent corners of the internet, and returns 10 candidates with sources, each already scored for pain, budget, reachability and speed.
---

# HUNT — ideas are not imagined, they are found in other people's complaints

An idea you thought of in the shower has no evidence attached. An idea you found because 400 people are angry about it in public, and 6 companies already charge for a bad version, comes with its market pre-attached. **This skill only produces the second kind.**

The output of a hunt is never "here are some ideas." It is **10 candidates, each with a URL, a quote, a price someone already pays, and a score.**

## BANNED (fake research — none of this counts as a hunt)
- **Ideas with no URL.** If you can't link to a human complaining, a competitor charging, or a number, you invented it. Delete it.
- **"AI for X" as a thesis.** Naming a technology is not naming a customer. Who bleeds, how much, and what do they pay today?
- **Trend-chasing without a buyer.** "Everyone's talking about it" is not "someone will send money."
- **Ideas that need a behaviour change.** The winner is usually a thing people already pay for, badly. Painkillers, not vitamins.
- **Markets you can't reach.** A perfect product for enterprise CIOs, sold by someone with no enterprise access, is a fantasy. Reachability is part of the score.
- **Copying a funded startup's pitch deck.** Their moat is capital; yours would be nothing. Find their unserved edge instead — the segment too small for them and plenty big for you.
- **Stopping at 3 candidates.** Ten, minimum. The tenth is usually the good one; you only find it because the first nine exhausted the obvious.
- **Only searching in English.** Thai, Japanese, Indonesian, Portuguese complaints are less picked over and cheaper to win.

## The hunt loop (run the tools — this is not a thinking exercise)
1. **Pick 5 hunting grounds** from [references/hunting-grounds.md](references/hunting-grounds.md), biased to where the operator already has an unfair advantage — their industry, their language, their city, their network.
2. **Search for pain in its native vocabulary**, not for "startup ideas." People in pain never write "painpoint." They write:
   - "is there a tool that", "how do you guys handle", "still doing this manually"
   - "I hate that", "why is X so expensive", "workaround for", "alternative to X"
   - "spreadsheet for", "template for", "someone I can pay to"
   - "we've been using X for years and it's still" · "cancelled my subscription because"
3. **Follow the money, not the noise.** For each candidate, find: who charges for it today, at what price, and how badly reviewed they are. A crowded market with 3-star incumbents is the best news in this document.
4. **Score every candidate** against the six-axis rubric in [references/signal-vs-noise.md](references/signal-vs-noise.md).
5. **Kill on the spot** anything that fails a hard filter (no budget holder, no reachable channel, illegal, licence-gated beyond reach).
6. **Write all 10 to `venture/candidates.md`** with their sources and scores — including the rejects and why. That file is your salvage yard on pivot day.
7. **Hand the survivors to [../market-gate/SKILL.md](../market-gate/SKILL.md).** Nothing gets committed to before it clears the ฿1B ceiling test.

## What a real candidate looks like
```
CANDIDATE 04 — invoice chasing for Thai SME accounting firms
PAIN      Bookkeepers manually chase 40–120 unpaid invoices/month by LINE, one by one.
SOURCE    <url to the thread> — "ทำมือทุกเดือน เสียเวลามาก" (2026-03-11, 87 replies)
          <url to a competitor's pricing page> — ฿2,900/mo, 2.9★ on reviews, no LINE support
PAYS NOW  ฿0 in software, ~14 h/month of a ฿25k/mo salary ≈ ฿2,200/mo of burnt time
BUYER     The firm's owner. Signs alone, no procurement, decides in one call.
REACH     3 Facebook groups (12k, 8k, 5k members) + a directory of ~4,000 registered firms
WHY NOW   e-Tax mandates pushed these firms onto digital records in the last 18 months
CEILING   → hand to market-gate
```
Anything less specific than this is not a candidate, it's a mood.

## Where the money hides (the seven veins)
1. **Manual work at scale** — people paying humans to do a repetitive thing. Every "we do it in a spreadsheet" is a product.
2. **Angry customers of an incumbent** — a 2-star review is a spec sheet with a budget attached.
3. **Broken handoffs between two tools** — the gap between two apps nobody owns.
4. **Regulation with a deadline** — new rules force spending on a timer. Deadlines beat features.
5. **Price gouging** — one dominant vendor, fat margins, hated. Undercut at 1/3 price with 1/10 of the features.
6. **Boring industries with old software** — the UI is from 2004 and the buyer is 55 and pays on time.
7. **A shift that just happened** — new platform, new law, new API, new cost, new behaviour. First one to sell into it wins twelve months of unopposed distribution.

## Speed rules
- A hunt is **one working session**, not a week of research. Ten candidates, sourced, scored, written down.
- **Depth after the gate, not before.** Do not fall in love with candidate #2 and spend three hours on it. Score all ten first.
- If none of the ten clears the ฿1B gate, do not lower the gate. **Change the hunting grounds** — you searched where the crowd searches.

## Done check
- [ ] 10 candidates, each with at least one source URL and one real quote
- [ ] Each names a specific buyer with the authority to pay, and what they pay today
- [ ] Each names a reachable channel the operator can actually use this week
- [ ] All six axes scored; hard filters applied
- [ ] All 10 written to `venture/candidates.md`, rejects included with reasons
- [ ] Survivors handed to market-gate — nothing committed to yet