💡 Bottom line first: startups no longer need "gut feeling" or random sales lists to find their first customers. With the Codex First Customer Finder Skill, you can gather evidence-backed "pain, demand, and timing" signals from public sources and get a prospect shortlist plus personalized opening messages in minutes. And it's 100% open source and free.

The God-Tier Tool for Startups Finding Their "First Customer" — Automate Lead Generation with a Codex Skill (with Japan Market Customization)

"The product is done. But where are the first customers?"

This is the wall every startup founder hits earliest. Scroll through social media, exchange business cards — you still rarely meet someone who "has this exact problem right now."

Enter the Codex First Customer Finder Skill. It's a skill you install into OpenAI's coding agent Codex — just give it your service URL or idea, and it lists evidence-backed prospects and even writes high-open-rate opening lines.

In this article, we explain everything from how the tool works, to mastering it for the Japanese market (Japanese, X lang:ja, note.com, etc.), to practical tips with Hermes Agent.

🛠 Tool Overview: What It Can Do

This skill reads your startup's URL, repository, or product description and automatically:

Defines your ICP (Ideal Customer Profile) — clarifies primary and adjacent customer profiles
Searches public signals — uncovers explicit "demand, pain, workaround, switch, timing" indicators
Builds a prospect list — evidence-based scoring, shortlisting promising candidates
Scores — scores fit and timing
Personalized openers — generates polite, source-based initial outreach messages
HTML report output — creates a responsive standalone HTML report

Installation (npx command)

npx --yes codex-first-customer-finder-skill@latest

That's it — it installs to ~/.codex/skills/first-customer-finder. Restart Codex and it's active.

Basic Usage Example

Use $first-customer-finder to find ten evidence-backed potential
first customers for https://example.com and create the final HTML report.

100% Open Source

Fully open source under the MIT license. Zero cost, and you can fork and customize it to your liking.

⚡ Key Features and Strengths

What makes this skill different from "just scraping" is its design philosophy.

Evidence-based — every prospect links to the public source it came from
Public data only — no private contact harvesting or sensitive personal data
Manual outreach premise — no automatic sending. It only assists with "lists and copy"

Available Modes

Mode	Description
quick	Up to 5 strong candidates
standard	Up to 10 from multiple sources
deep	Up to 20 + iterative pattern analysis
design-partners	Feedback-oriented early adopters
b2b	Companies with public business triggers
community	Explicit requests and public discussion signals

🇯🇵 How to Use It in the Japanese Market

It's strong out of the box, but customizing for the Japanese market boosts accuracy dramatically.

Japan-Specific Customization Examples

Prioritize Japanese search — add "Japanese," "Japanese SaaS," "domestic startups" to search queries
Use X lang:ja — pick up Japanese posts like "lang:ja 困ってるツール探してる" (struggling, looking for a tool)
Add note.com / Qiita as sources — where Japanese engineers and entrepreneurs talk about their problems

Additional Prompt Examples

Use $first-customer-finder in b2b mode for [URL].
Prioritize Japanese public signals (X lang:ja, note.com, Qiita) and
list departments/owners at Japanese B2B companies that mention
"属人化" (tribal knowledge), "手作業" (manual work), or "レガシーシステム" (legacy systems).

Use $first-customer-finder in design-partners mode for [URL].
Prioritize individuals in Japanese developer communities (Zenn, Qiita)
who publish OSS similar to our product.

Advice Considering Japanese Business Culture

Japanese B2B places heavy weight on relationship building (trust), so cold outreach out of nowhere tends to be avoided. Use the skill's "polite openers with sources" as a foundation — start with likes and comments on X or in communities, build the relationship, then move to DM or email. That's the realistic path.

Effectiveness Assessment

Very effective as a supporting tool. Alone, though, it's limited — ultimately "human relationship building" is what decides. Use it correctly: to narrow down "who to reach out to."

✅ Pros and Cons

Pros

Free — MIT-licensed OSS, zero running cost (Codex API usage separate)
Fast — pass a URL and get a report in minutes
High-quality personalization — source-based, so it doesn't feel spammy
Strong for market validation — get "demand backing" from existing public signals

Cons

Language dependent — Japanese support requires prompt-side tweaks
Cultural limits — Japan's relationship-building process can't be automated
Depends on public signal volume — too-niche problems are harder to find

🤖 Practical Tips with Hermes Agent (or Codex)

Japan Market Prompt Example (reusable in Hermes too)

When calling Codex from Hermes Agent's terminal:

cd ~/lead-research && codex exec --full-auto \
"Use \$first-customer-finder in b2b mode for https://your-startup.example.
Prioritize Japanese signals (X lang:ja, note.com, Qiita) and list 20 prospects,
outputting the HTML report to first-customers.html."

How to Persist the Skill

Install manually so it's always available:

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

After restarting Codex, $first-customer-finder is available in any project.

📚 Related Articles (Dive Deeper)

AI Coding Agent Complete Comparison: Cursor vs Claude Code vs Codex vs Hermes vs Fugu — deep comparison of each agent
Hermes Agent Memory Comparison Guide — practical article generation and automation methods
Automatically Finding Your Startup's First Customers with AI (Codex CLI × Hermes Agent) — the full lead generation picture

🎯 Summary and Recommendation

The Codex First Customer Finder Skill is a god-tier tool that dramatically simplifies the first step of finding "first customers."

Recommended for

Startup founders who don't know "who to sell to"
SaaS developers looking for early adopters (design partners)
Entrepreneurs interested in the Japanese market, teams targeting global expansion

Next Actions

Try it now — run npx --yes codex-first-customer-finder-skill@latest
Fork and customize — add Japanese sources and adapt it for your company
Combine with Hermes Agent — automate content creation and follow-up from the list

Start by passing your product URL and see what "evidence-backed prospects" come back — try it today.

🔗 Reference Links

GitHub repository: Kappaemme-git/codex-first-customer-finder-skill
Original X post (related): find your first customers with codex
Codex CLI official: https://github.com/openai/codex
Hermes Agent official: https://hermes-agent.nousresearch.com/

The Codex First Customer Finder Skill introduced in this article is MIT-licensed open source and contains no affiliate links from which the author earns money (a purely productivity-focused introduction).