| name | codex-first-customer-finder-skill |
| description | Find evidence-backed potential first customers from public signals for startups using Codex AI |
| triggers | ["find first customers for my startup","research potential early adopters","discover design partners from public signals","find b2b prospects with timing signals","generate customer discovery report","identify early customers with evidence","search for startup prospects with demand signals","create first customer shortlist"] |
Codex First Customer Finder Skill
Skill by ara.so — Codex Skills collection.
A Codex skill that analyzes startup URLs or product ideas to find qualified potential first customers using recent public pain, demand, and timing signals. It creates evidence-backed prospect shortlists with source links, fit scores, and personalized outreach openers without automatically sending any messages.
What It Does
- Analyzes startup URLs, repositories, or product descriptions to define ideal customer profiles
- Searches public sources for explicit demand, pain, workaround, switching, and timing signals
- Qualifies prospects with evidence-based scores linking to original public sources
- Drafts respectful, source-based outreach openers (manual sending only)
- Generates responsive standalone HTML reports with prospect rankings
- Avoids private contact enrichment and sensitive personal data collection
Installation
The skill installs into ~/.codex/skills/first-customer-finder:
npx --yes codex-first-customer-finder-skill@latest
After installation, restart your Codex agent to load the skill.
Manual Installation
git clone https://github.com/Kappaemme-git/codex-first-customer-finder-skill.git
mkdir -p ~/.codex/skills
cp -R codex-first-customer-finder-skill/first-customer-finder ~/.codex/skills/first-customer-finder
Usage Patterns
Basic Customer Discovery
Find potential first customers for a startup:
Use $first-customer-finder to find ten evidence-backed potential first customers for https://example.com and create the final HTML report.
Design Partners Mode
Prioritize prospects likely to provide product feedback:
Use $first-customer-finder in design-partners mode for this startup: https://example.com. Prioritize people publicly describing the problem and likely to give product feedback.
B2B Research
Find business prospects with public triggers:
Use $first-customer-finder in b2b mode for https://example.com. Find public business triggers, qualify the relevant companies, and draft one opener per prospect without sending anything.
Community Signal Discovery
Focus on explicit requests and public discussions:
Use $first-customer-finder in community mode for [product description]. Find people actively discussing this problem in public forums and communities.
Quick Research
Generate a shortlist of high-confidence prospects:
Use $first-customer-finder in quick mode for https://example.com to find five strong prospects with the best timing signals.
Deep Analysis
Comprehensive research with pattern analysis:
Use $first-customer-finder in deep mode for https://example.com. Find up to twenty prospects and analyze repeated pain patterns across signals.
Modes
| Mode | Description | Prospect Count |
|---|
quick | High-confidence prospects only | Up to 5 |
standard | Balanced research across source types | Up to 10 |
deep | Comprehensive with pattern analysis | Up to 20 |
design-partners | Feedback-oriented early adopters | Up to 10 |
b2b | Companies with business triggers | Up to 10 |
community | Public discussion signals | Up to 10 |
Report Output
The generated HTML report includes:
- Early-customer verdict - Overall market signal assessment
- Primary ICP and disqualifiers - Ideal customer profile definition
- Highest-confidence prospect - Top-ranked opportunity with evidence
- Evidence-backed prospect shortlist - Complete list with source links
- Fit and timing scores - Quantified prospect qualification
- Source links and signal dates - Original public signal references
- Personalized outreach openers - Draft messages (manual sending)
- Repeated pain patterns - Common themes across signals
- Seven-day manual outreach plan - Suggested sequencing
- Research limitations - Methodology transparency
Configuration
The skill uses environment variables for API access (if needed for data sources):
export CODEX_SKILL_DATA_SOURCES="public"
export CODEX_SKILL_MAX_PROSPECTS=10
export CODEX_SKILL_OUTPUT_DIR="./customer-reports"
Invocation Examples
From Product Description
Use $first-customer-finder for a developer tool that helps teams migrate from MongoDB to PostgreSQL. Focus on recent migration pain signals and active database switchers.
From Repository
Use $first-customer-finder to analyze https://github.com/username/project and find ten potential first customers based on the README and documentation.
With Specific Criteria
Use $first-customer-finder in standard mode for https://example.com. Focus on prospects who have publicly mentioned the problem in the last 30 days and show urgency signals.
Understanding Prospect Scores
- Fit Score: How well the prospect matches the ICP (based on role, context, problem description)
- Timing Score: Recency and urgency of the public signal
- Evidence Quality: Clarity and specificity of the pain point or demand signal
High scores (8-10) indicate strong alignment between the prospect's public signal and your product's value proposition.
Best Practices
1. Start with the Startup Context
Provide as much context as possible:
Use $first-customer-finder for https://example.com (a scheduling tool for freelance consultants). Look for signals around calendar chaos, double-booking, and manual scheduling pain.
2. Review Evidence Links
Always verify the source links in the report before reaching out. The skill provides hypotheses, not guarantees.
3. Customize Outreach Openers
The generated openers are templates. Personalize them further based on:
- Additional profile research
- Specific signal context
- Your brand voice
4. Respect Privacy and Intent
- Signals are public, but respect boundaries
- Only reach out if the signal indicates openness to solutions
- Never automate bulk outreach
5. Track Follow-up Manually
The skill generates a 7-day plan but doesn't send messages. Use your CRM or manual tracking for:
- Initial outreach status
- Response tracking
- Follow-up scheduling
Troubleshooting
Skill Not Found
If Codex doesn't recognize $first-customer-finder:
ls ~/.codex/skills/first-customer-finder
npx --yes codex-first-customer-finder-skill@latest
Low-Quality Prospects
If prospects seem irrelevant:
Use $first-customer-finder for https://example.com. Define the ICP more narrowly: exclude hobbyists, focus on B2B SaaS companies with 10-50 employees experiencing [specific pain point].
Insufficient Signal Volume
Try different modes or sources:
Use $first-customer-finder in deep mode for https://example.com. Expand sources to include Reddit, niche forums, and GitHub discussions beyond Twitter/LinkedIn.
Report Not Generating
Ensure the skill completes research before requesting the report:
Use $first-customer-finder for https://example.com in standard mode. After completing prospect research, generate and save the HTML report to ./reports/customers.html
Integration Workflows
With CRM Export
Use $first-customer-finder for https://example.com. After generating the report, extract prospect data (name, signal, source URL, score) into a CSV for CRM import.
With Content Calendar
Use $first-customer-finder in community mode for https://example.com. Identify the top three pain patterns and suggest five content ideas addressing each pattern for inbound customer acquisition.
With Product Roadmap
Use $first-customer-finder in design-partners mode for https://example.com. Find prospects who mention feature gaps or workarounds that could inform our Q2 roadmap priorities.
Limitations
- Public signals only: No private data scraping or contact enrichment
- Hypotheses, not guarantees: Prospects are qualified leads, not confirmed buyers
- Manual outreach required: Skill never sends messages automatically
- Source availability: Results depend on public signal volume and recency
- Language: Primary support for English-language signals
Example Complete Workflow
1. Use $first-customer-finder in standard mode for https://myproduct.com (a Git-based documentation tool for engineering teams)
2. After research completes, review the HTML report focusing on:
- Top 3 prospects with timing scores above 8
- Common pain patterns in the "Repeated Patterns" section
- Source links for signal verification
3. Customize the suggested outreach openers for the top 3 prospects, adding:
- Reference to their specific project or company
- One relevant product feature addressing their exact pain point
- Low-pressure ask (demo vs. feedback vs. conversation)
4. Manually send outreach via LinkedIn/email with 2-day spacing
5. Use the 7-day plan to schedule follow-ups in calendar
6. Track responses and update ICP assumptions based on feedback
License
MIT License - See repository for full details.
