Prior-Art Search · Exhaustive Similar Projects

When to Trigger

• After idea-eval runs (regardless of 🟢🟡🔴 — finding an existing solution = zero-effort delivery)
• Founder says "has someone already built this idea"
• Dashboard pipeline Stage 2 triggers

Input

• docs/ideas/<slug>/eval.md (idea summary + user persona + pain points)

Search Strategy · 8 Angles (must search at least once per angle)

At least 5 WebSearch calls — fewer means you're cutting corners.

Evaluate Each Candidate

For every project found (GitHub / product / open-source library / service), output:

Output File

docs/ideas/<slug>/prior-art.md:

---
stage: prior-art
slug: <slug>
aiVersion: 1
generatedAt: 2026-04-26
generatedBy: [email protected]
candidates: 8                 # total found
exactMatch: false             # is there one that's basically what you want
recommendedPath: "fork-modify" # use-as-is / fork-modify / borrow-pattern / build-fresh
overallVerdict: "🟡 Existing similar solution exists. Forking is 5-10x faster than building from scratch."
sources:
  - github
  - hackernews
  - twitter
  - jike
  - awesome-list
userEdits: {}
---

# Similar Projects Research · <slug>

## Overall Assessment

[1-3 paragraphs: whether a similar solution was found, recommended path, rationale]

## Candidates (sorted by overlap descending)

### 1. <Project Name> · <Overlap X%>

- **URL**: [link]
- **Stars / Status**: 12.3k · actively maintained
- **Summary**: ...
- **Overlap**: [what it does that you want]
- **Gap**: [what it doesn't do that you want]
- **Source**: direct search / [big-name recommendation link]
- **Recommendation**: A use-as-is · just add 1 small filter
- **Cost**: 0 lines of code vs 8 weeks to build

### 2. ...

## Integration Recommendations

[If fork-modify path: give specific modification steps + time estimate]
[If use-as-is path: give registration/deployment steps]
[If build-fresh path: explain why none of the 8 candidates suffice and what unique value justifies building from scratch]

## Decision Criteria for 4 Paths

| Path | Trigger | Action |
|---|---|---|
| **A. Use-as-is** | Overlap ≥ 90% · actively maintained · acceptable license | Register / subscribe / install · trial for 7 days · revisit if insufficient |
| **B. Fork-modify** | Overlap 60-89% · open source · moderate maintenance | Clone · get it running · modify 1-2 key gaps · estimate 1-3 days |
| **C. Borrow-pattern** | Overlap 30-59% · idea is right but implementation is too far off | Read source code to learn architecture · reuse architectural decisions when building |
| **D. Build-fresh** | Overlap < 30% or all candidates are dead / closed-source and unmodifiable | Only then enter requirements phase |

## Solo Founder Principles

- **Finding = winning**: Not embarrassing — it saves 8 weeks of runway
- **Always search before building**: 90% of "I have a new idea" has already been built somewhere in the world
- **Big-name recommendations matter**: Their tools may not have many stars, but quality has been vetted by them

## Prohibited

- ❌ Cutting corners with only 1-2 searches (must use ≥5 angles)
- ❌ Giving "recommendations" without specific URLs (not verifiable)
- ❌ Labeling all candidates as "< 30% overlap" to force build-fresh (this is founder optimism bias — stay neutral)
- ❌ Missing Chinese communities (Jike / Xiaohongshu / V2EX / Zhihu) — Chinese circles have many undervalued tools

## Collaboration with Other Tools

- Input comes from `idea-eval`
- Output feeds into `requirements` (if fork-modify path, requirements only list the gaps)
- Works with `validator`: validator will ask "why not use X" — this step answers that and avoids review

## 4 Recommended Path Decision Criteria

| Path | Trigger | Action |
|---|---|---|
| **A. Use-as-is** | Overlap ≥ 90% · actively maintained · acceptable license | Register / subscribe / install · trial for 7 days · revisit if insufficient |
| **B. Fork-modify** | Overlap 60-89% · open source · moderate maintenance | Clone · get it running · modify 1-2 key gaps · estimate 1-3 days |
| **C. Borrow-pattern** | Overlap 30-59% · idea is right but implementation is too far off | Read source code to learn architecture · reuse architectural decisions when building |
| **D. Build-fresh** | Overlap < 30% or all candidates are dead / closed-source and unmodifiable | Only then enter requirements phase |

## Solo Founder Principles

- **Finding = winning**: Not embarrassing — it saves 8 weeks of runway
- **Always search before building**: 90% of "I have a new idea" has already been built somewhere in the world
- **Big-name recommendations matter**: Their tools may not have many stars, but quality has been vetted by them

## Prohibited

- ❌ Cutting corners with only 1-2 searches (must use ≥5 angles)
- ❌ Giving "recommendations" without specific URLs (not verifiable)
- ❌ Labeling all candidates as "< 30% overlap" to force build-fresh (this is founder optimism bias — stay neutral)
- ❌ Missing Chinese communities (Jike / Xiaohongshu / V2EX / Zhihu) — Chinese circles have many undervalued tools

## Collaboration with Other Tools

- Input comes from `idea-eval`
- Output feeds into `requirements` (if fork-modify path, requirements only list the gaps)
- Works with `validator`: validator will ask "why not use X" — this step answers that and avoids review