Browse categories
Explore
Fiverr Pro
English
$
USD
I will set up claude code for your team with custom skills, hooks, and mcp
Shion nagaya
About this gig
Your team installed Claude Code. Now make it actually work.
I'm a QA director at a Japanese IT services company where Claude Code runs in production daily generating tests, reviewing code, automating workflows. I hold all 4 Anthropic certifications (Claude API, Claude Code, Agent Skills, MCP), and I set teams up so AI output is consistent, safe, and fast.
What I deliver:
A CLAUDE.md written for YOUR codebase conventions the AI must follow
Permission & settings design, safe by default
Custom Skills and Hooks for your recurring workflows
MCP connections (GitHub, databases, internal tools)
A team playbook + Loom walkthroughs no meetings needed
Why async works: everything ships as files, docs, and recordings your team keeps forever. Questions answered in writing within hours (JST).
How it works:
1. You share your repo structure and goals (read-only or a sanitized sample is fine)
2. I design and deliver the setup with documentation
3. Advanced tier adds one working automation + 14 days of async support
Message me with your stack and what you want Claude Code to do concrete plan within 12 hours.
Get to know Shion nagaya
Shion nagaya
QA Director,AIDriven Test Automation
- FromJapan
- Member sinceSep 2026
Languages
Japanese, English
I help teams ship quality software by combining QA discipline with modern AI tooling.
5 years in QA & development: 3+ years at Japan's largest QA firm (TSE Prime) managing test teams for major game publishers, then dev team lead, now Director of QA at an IT services company in Japan.
What I offer:
- AI-driven test automation: Playwright + Claude Code (4 Anthropic certs)
- Rigorous manual QA: test planning, edge cases, clear bug reports
- Japan-based: native Japanese, JP payments & market compliance
Fully remote (JST), async & text-first. On-time, no surprises.
My Portfolio
FAQ
Do I need to give you access to our private code?
Not necessarily. Read-only access produces the best results, but a sanitized repo structure plus a few representative files works well too. I'm happy to sign an NDA before you share anything.
What are Skills, Hooks, and MCP in plain terms?
Skills = reusable instructions Claude follows for specific tasks (e.g., "write tests our way"). Hooks = automatic checks that run around the AI's actions. MCP = connectors that let Claude safely use your tools (GitHub, databases, docs).
Which package should I choose?
Solo dev or one repo: Solid Foundation. A team that wants consistent AI usage with custom workflows: Team Enablement. If you also want one automation working end-to-end (e.g., PR review or spec-to-test generation): Full Workflow Build.
Does this work for non-engineering teams?
Yes. The same setup approach works for document-heavy workflows (reports, research, operations). Tell me your use case and I'll confirm fit before you order.
How do we communicate?
Async and text-first: written updates, documentation, and Loom recordings. I'm based in Japan (JST / UTC+9) and respond within a few hours during JST daytime and evenings. No live meetings needed.
What stacks and languages do you support?
TypeScript/JavaScript, Python, and most common web stacks. The setup approach (conventions, skills, permissions, MCP) is stack-agnostic — tell me yours and I'll confirm before you order.
Who owns the deliverables?
You do, 100%. All configuration, skills, documentation, and recordings are delivered to your repository and remain yours. No lock-in.
What happens when Claude Code gets updated?
Deliverables follow current best practices at delivery time. Within the Full Workflow Build tier's 14-day support window I update anything affected by changes. Ongoing maintenance is available as a follow-up order.
Can you train our team live?
I work async by design — instead of live sessions you get a written playbook, Loom walkthroughs, and fast written Q&A. Teams usually find this better: nothing is lost after the call, and every answer stays searchable.
Is our data safe?
Yes. We agree on what the agent may and may not touch, permissions are configured accordingly, and secrets never go into prompts or configs. I can follow your security guidelines — share them before we start.
