사용 방법
- 설치 명령을 복사하고 자신의 에이전트 환경에 설치하기 전에 소스 패키지를 확인하세요.
- 필요한 도구, 인증 정보, 지원 환경은 원본 문서에서 확인하세요.
- 입력과 원하는 산출물을 설명하고 출처와 결과를 대조하세요.
npx skills add https://github.com/solanabr/solana-ai-kit --skill idea-sprint시작 전 확인
- 설치에는 모델 구독, API 한도, 파일 접근 권한이 포함되지 않습니다.
- 별점 수와 설치 수는 활동 지표이며 독립적인 품질·보안 검증이 아닙니다.
원본 저장소의 제목, 예시, 코드, 표, 링크, 이미지를 유지해 표시합니다.
<!-- Adapted from sendaifun/solana-new (find-next-crypto-idea, validate-idea), MIT © 2026 SendAI and Superteam. Telemetry removed. -->
Idea Sprint
Interview → necessity gate → 3 candidates → score → go/no-go. Output is a decision, not a brainstorm.
Context handoff
- At start: read
.claude/context/idea.mdand.claude/context/build.mdif present — resume from prior state instead of re-interviewing. - On completion: write/update
.claude/context/idea.mdwith the chosen idea, scores, validation evidence, and open risks. Downstream skills (pitch-deck, hackathon) read it.
Workflow
1. Blunt interview
No flattery. Short, pointed questions, one at a time, until three things are explicit:
- Edge — what the founder knows/can do that most can't (domain, distribution, tech)
- Constraint — time, money, team, chain commitments
- Wedge — the niche entry point, not the end-state vision
Push back on vague answers. "DeFi for everyone" is not a wedge. Full question bank: interview-framework.md.
2. Crypto-necessity gate
Kill question: "What gets worse if I remove the blockchain?" If the answer is vague, aesthetic, or marketing-driven — redirect the idea before scoring it. Pass criteria and redirect patterns: crypto-necessity-test.md.
3. Exactly 3 candidates
Generate three diverse candidates (different mechanisms/markets, not three flavors of one idea). For each:
- One-line pitch + target user
- Winner case — what's true in 18 months if it works
- Bear case — the most likely way it dies
Seed from datasets + live landscape (below), then combine with fresh research. Datasets are inspiration, not constraints.
4. Score /15
Each candidate, 0–3 per dimension (full anchors: scoring-rubric.md):
| Dimension | 3 means |
|---|---|
| Founder fit | unfair advantage |
| MVP speed | shippable in under a week |
| Distribution | first ten users are obvious |
| Market pull | people already paying for bad alternatives |
| Revenue path | clear monetization story |
5. Validate + go/no-go
Check demand signals against customer-signal-rubric.md — manual workarounds, active forks, bounties, on-chain activity = real; likes and "cool idea" replies = noise. Sprint structure: validation-framework.md.
- ≥ 8/15 → go. Write
idea.md, suggest/scaffoldnext. - 6–7 → conditional: name the one dimension to de-risk first.
- < 6 → strong no-go. Every no-go gets a pivot suggestion — use pivot-or-persist.md.
6. Write .claude/context/idea.md
Chosen idea, wedge, scores table, demand evidence, bear case, next step.
Idea datasets (inert JSON, ~515 entries)
In ../ext/solana-new/skills/data/ideas/:
- web3-ideas-combined.json — master list (summary)
- a16z-big-ideas-2025.json, a16z-state-of-crypto-2025.json
- yc-requests-for-startups.json, yc-crypto-companies.json
- alliance-ideas.json, superteam-ideas.json
- rwa-defi-2026-ideas.json, yash-defi-2024-ideas.json
Idea-source guides (markdown commentary on the same sources): ../ext/solana-new/skills/data/guides/ — plus source-map.md and research-playbook.md for where/how to research live.
Live hackathon landscape
ext/colosseum — 5,400+ Colosseum submissions for crowdedness checks, winner patterns, and gap analysis (requires COLOSSEUM_COPILOT_PAT).
Output format
Report spec (3 ranked candidates, scores, decision): output-spec.md.
