Demand Discovery Skill

中文版

Automated niche demand discovery for indie developers. Scans 7 data sources across 26 categories to find real, underserved demands that are too small for companies but perfect for solo builders.

What It Does

Runs a structured 9-stage research workflow:

1. Preparation — Pick next category from 26-category rotation
2. Reddit — Mine pain points from subreddits
3. G2 — Analyze competitor low-star reviews
4. Product Hunt — Discover trending alternatives
5. GitHub Issues — Find feature requests with high reaction counts
6. Hacker News — Analyze developer discussions
7. AlternativeTo — Map software replacement gaps
8. Google Trends — Validate search volume trends
9. Synthesis — Score demands, generate report

Each demand gets:

• Volume Score (0-100) — Weighted composite from 7 sources
• Trend — Rising / Stable / Declining
• Feasibility (1-5) — How doable for an indie developer

Data Sources

MCP Server	Data Source	Auth Required
niche-reddit-mcp	Reddit API	OAuth2 (free)
niche-g2-mcp	G2.com (browser)	None
niche-producthunt-mcp	Product Hunt API	Token (free)
niche-github-issues-mcp	GitHub API	Token (optional)
niche-hackernews-mcp	HN Algolia API	None
niche-alternativeto-mcp	AlternativeTo	None
niche-google-trends-mcp	Google Trends	None

All 7 servers are Python packages, auto-installed via uvx. Missing sources are skipped gracefully — scoring weights redistribute automatically.

Supported Platforms

Platform	Install	Trigger	MCP Support
Claude Code	install.sh (auto)	/demand-discovery	Native
Gemini CLI	install.sh (auto)	/demand-discovery	Via shell
OpenCode / MiMo-Code	install.sh (auto)	/demand-discovery	Native
Hermes	install.sh (auto)	Say "demand discovery"	Via shell
Codex CLI	Copy AGENTS.md	Say "demand discovery"	Via shell
Cursor	Copy .cursor/rules/	Say "demand discovery"	Via web search
Windsurf	Copy .windsurfrules	Say "demand discovery"	Via web search
GitHub Copilot	Copy .github/	Say "demand discovery"	Via web search
Cline / Roo Code	Copy .clinerules	Say "demand discovery"	Via shell
Continue.dev	Copy .continuerules	Say "demand discovery"	Via shell
Aider	Copy .aider.conf.yml	/run demand-discovery	Via shell
Trae	Copy .trae/rules/	Say "demand discovery"	Via web search
Augment Code	Copy AGENTS.md	Say "demand discovery"	Via shell
iFlow	install.sh (auto)	/demand-discovery	Native
Any AGENTS.md-compatible agent	Copy AGENTS.md	Say "demand discovery"	Varies

Quick Install

One-liner (auto-detects all installed platforms):

git clone https://github.com/Ddhjx-code/Demand-Discovery-Skill.git
cd Demand-Discovery-Skill && ./install.sh

Claude Code only:

git clone https://github.com/Ddhjx-code/Demand-Discovery-Skill.git \
  ~/.claude/skills/demand-discovery

Gemini CLI only:

git clone https://github.com/Ddhjx-code/Demand-Discovery-Skill.git
cp Demand-Discovery-Skill/.gemini/commands/demand-discovery.toml ~/.gemini/commands/

Cursor / Windsurf — copy into your project:

git clone https://github.com/Ddhjx-code/Demand-Discovery-Skill.git
# Cursor
cp -r Demand-Discovery-Skill/.cursor/rules/ your-project/.cursor/rules/
cp Demand-Discovery-Skill/SKILL.md your-project/
cp -r Demand-Discovery-Skill/references/ your-project/references/
# Windsurf
cp Demand-Discovery-Skill/.windsurfrules your-project/
cp Demand-Discovery-Skill/SKILL.md your-project/
cp -r Demand-Discovery-Skill/references/ your-project/references/

Any AGENTS.md-compatible agent (Codex, Hermes, Augment, etc.):

cp Demand-Discovery-Skill/AGENTS.md your-project/
cp Demand-Discovery-Skill/SKILL.md your-project/
cp -r Demand-Discovery-Skill/references/ your-project/references/

Usage

/demand-discovery              # auto-pick next category
/demand-discovery education    # specific category

Or say: "run demand discovery", "niche research", "find indie opportunities"

Output

Each run produces (in the skill's output/ directory):

• Daily report — output/reports/YYYY-MM-DD-<category>.md
• Findings index — Cumulative ranked list at output/findings-index.md
• Tracker — Rotation state at output/tracker.md

Sample Report

#	Demand	Category	Evidence	Score	Trend	Feasibility
1	AI Agent context management	Dev Tools	HN: 1318pts; GH: 5585 reactions	82	Rising	4
2	Terraform config preprocessor	IaC	GH: 7000+ reactions across 5 issues	75	Stable	3

Category Rotation (26)

#	Category
1	Developer Tools & CLI
2	Small Business Ops (invoicing, scheduling)
3	Content Creation & Writing
4	Design & Prototyping
5	E-commerce & Shopify
6	Education & Learning
7	Health & Fitness
8	Personal Finance
9	Recruiting & HR
10	SEO & Marketing
11	Data Visualization
12	API Tools & Integrations
13	Remote Work
14	Productivity & Time Tracking
15	Legal & Compliance
16	Real Estate
17	Food & Restaurant
18	Social Media Management
19	Customer Support
20	Analytics & A/B Testing
21	Email & Newsletter
22	File Conversion & Docs
23	Browser Extensions
24	AI Wrappers & Prompts
25	Accessibility
26	Vertical SaaS

Credential Setup

Reddit (free)

1. Go to https://www.reddit.com/prefs/apps
2. Create a "script" type app
3. Copy client_id (under app name) and client_secret

Claude Code / iFlow — add to ~/.claude/settings.local.json:

{
  "mcpServers": {
    "niche-reddit-mcp": {
      "env": {
        "REDDIT_CLIENT_ID": "your_id",
        "REDDIT_CLIENT_SECRET": "your_secret"
      }
    }
  }
}

Other platforms — set environment variables:

export REDDIT_CLIENT_ID="your_id"
export REDDIT_CLIENT_SECRET="your_secret"

Product Hunt (free)

1. Go to https://www.producthunt.com/v2/oauth/applications
2. Create an app, get the Developer Token
3. Set PRODUCTHUNT_TOKEN via settings or env var

GitHub (optional, increases rate limit)

1. Go to https://github.com/settings/tokens
2. Create Personal Access Token with public_repo scope
3. Set GITHUB_TOKEN via settings or env var

G2 (optional login for full reviews)

uvx niche-g2-mcp --login

Requirements

• Python >= 3.11
• uv (for uvx command)
• Chrome (for G2 browser automation, optional)

License

Apache-2.0