What if a project manager could spin up a production-grade Hyperledger Fabric network by typing "create a Fabric testnet with two orgs" into a terminal? No YAML files, no Docker Compose debugging, no crypto material generation. Just a sentence.
That's not a hypothetical. It's what happens when you combine Claude Code's skill system with ChainLaunch's blockchain infrastructure platform. We built six custom skills that teach Claude Code exactly how to provision, manage, and troubleshoot blockchain networks — and the results have changed how our team and early adopters work with enterprise blockchain.
This post walks through what Claude Code skills are, how we built ours, and how you can try the entire workflow yourself in under 10 minutes — even if you've never touched blockchain infrastructure before.
TL;DR: Claude Code skills are an open standard (works in Claude Code, Codex CLI, Cursor, and Gemini CLI) that teach AI assistants domain-specific workflows. ChainLaunch ships 6 skills covering CLI operations, deployment, code review, changelogs, and documentation — letting anyone deploy Hyperledger Fabric or Besu networks from plain English. One command provisions a full VPS:
curl -fsSL https://chainlaunch.dev/deploy.sh | bash.
What Are Claude Code Skills, and Why Do They Matter?
84% of developers are now using or planning to use AI coding tools — up from 76% in 2024 (Stack Overflow Developer Survey, 2025). But adoption doesn't mean effectiveness. The same survey found that 46% of developers actively distrust AI tool output accuracy. Only 3.1% "highly trust" AI-generated code.
That trust gap is where skills come in.
Claude Code skills are markdown files (named SKILL.md) that live inside your project's .claude/skills/ directory. Each skill gives Claude Code structured knowledge about a specific workflow — CLI flags, API patterns, deployment steps, review checklists — so it can execute complex tasks accurately without you explaining the same context every conversation.
Anthropic launched skills as an open standard in December 2025. The same SKILL.md format now works across Claude Code, OpenAI Codex CLI, Cursor, and Gemini CLI. You're not locked into one tool — skills travel with your codebase.
Think of them as persistent, version-controlled prompts. When someone on your team asks Claude Code to "deploy a Besu testnet with 4 validators," the chaindeploy-cli skill provides the exact command syntax, flag defaults, and post-deployment verification steps. Claude Code doesn't guess — it follows a documented playbook.
Why does this matter for blockchain infrastructure? Because the gap between "I want to test this idea" and "I have a running network" kills most enterprise blockchain projects. A Hyperledger Foundation survey found that infrastructure complexity is the top barrier to production for 73% of enterprise blockchain initiatives. The global blockchain market hit $32.99 billion in 2025 and is projected to reach $393.45 billion by 2030 (MarketsandMarkets, 2025) — but most of that growth depends on teams actually getting past the infrastructure phase.
Skills close that gap by encoding operational expertise into something any team member can invoke.
For teams already using ChainLaunch, our Besu network setup guide and Fabric deployment tutorial show the manual CLI workflow. Skills automate all of it.
What Skills Does ChainLaunch Ship?
We maintain six production skills across the monorepo. Each one targets a specific phase of the blockchain infrastructure lifecycle.
CLI Reference Skills
chaindeploy-cli — The core skill. It contains every command, subcommand, and flag for the ChainLaunch binary. When you ask Claude Code to create a Fabric network, it pulls the exact syntax from this skill:
chainlaunch testnet fabric --name mynet --org Org1 \
--peerOrgs Org1 --ordererOrgs Orderer1 \
--peerCounts Org1=2 --ordererCounts Orderer1=3 \
--mode dockerThe skill also documents the full Fabric network lifecycle — the six steps from network creation through chaincode deployment that teams typically get wrong. Claude Code follows them in order, every time.
chainlaunch-pro-cli — Extends the core CLI reference with enterprise features: RBAC commands, OIDC configuration, backup scheduling, multi-instance federation, and node sharing between organizations.
Operational Skills
deploy-chainlaunch — Guides production deployment from VPS provisioning through TLS configuration. Covers server sizing (2+ vCPU, 4GB+ RAM), Docker installation, SQLite setup, and multi-instance federation for Pro users.
go-review — A code review checklist that Claude Code applies when reviewing Pull Requests. Covers error handling patterns, SQLite/sqlc compliance, RBAC enforcement, API design conventions, and test coverage requirements. Every PR gets the same rigorous review, whether it's submitted at 2 PM or 2 AM.
write-changelog — Parses conventional commits (feat(scope): description), groups them by category, and produces formatted CHANGELOG entries. Release prep that used to take 30 minutes now takes seconds.
technical-docs — Orchestrates a multi-agent documentation pipeline. It delegates to four sub-agents (root creator, project creator, reference creator, and auditor) to produce architecture docs, API specs, database schemas, runbooks, and ADRs. One command, full documentation tree.
How They Work Together
Here's what a typical session looks like for a project manager evaluating blockchain for a supply chain use case:
You: Create a Fabric testnet with one org called SupplyChainOrg,
2 peers, and 3 orderers. Then deploy a supply chain chaincode.
Claude Code: [reads chaindeploy-cli skill]
[runs testnet fabric command]
[verifies nodes are running]
[sets anchor peers via API]
[pulls connection profile]
[writes Go chaincode]
[deploys via fabric install]
[runs test queries]
Done. Your network is running with the supply chain
chaincode deployed on channel "mychannel".
No config files edited. No documentation consulted. The skill provided the entire playbook.
How Do You Build a Claude Code Skill?
A skill is a markdown file with YAML frontmatter. Here's the anatomy of ChainLaunch's deploy-chainlaunch skill:
---
name: deploy-chainlaunch
description: >
Guides deployment of ChainLaunch to production servers.
Covers server provisioning, Docker installation, SQLite setup,
TLS configuration, and multi-instance federation.
---
# Deploy ChainLaunch
Step-by-step guide for deploying ChainLaunch to production.
## Prerequisites
- Server with 2+ vCPU, 4GB+ RAM
- Domain with DNS access
- SSH key pair
## Steps
### 1. Provision Server
...The name field becomes the skill's invocation handle. The description tells Claude Code when this skill is relevant. The body contains the actual instructions — commands, flags, verification steps, and gotchas.
Best Practices We Learned Building 6 Skills
1. Be prescriptive, not descriptive. Don't explain what a command does in theory. Show the exact command with real flags and realistic values. Claude Code performs better with concrete examples than abstract documentation.
# Bad
The testnet command creates a Fabric network with configurable parameters.
# Good
chainlaunch testnet fabric --name mynet --org Org1 \
--peerOrgs Org1 --ordererOrgs Orderer1 \
--peerCounts Org1=2 --ordererCounts Orderer1=3 \
--mode docker2. Include the full workflow, not just individual commands. Our chaindeploy-cli skill doesn't just list commands — it documents the 6-step Fabric network lifecycle end to end. Without this, Claude Code would create a network but skip anchor peer configuration, and chaincode discovery would silently fail.
3. Document what's NOT in the CLI. Some operations only exist as REST API calls. Our skill has a table of API-only operations (anchor peers, block queries, channel config) so Claude Code knows to use curl instead of looking for a CLI flag that doesn't exist.
4. Put verification steps after every action. After creating a testnet, the skill instructs Claude Code to check node status and block height. This catches failures immediately instead of 10 steps later.
### 2. Verify nodes are running
curl -u $USER:$PASS $API_URL/nodes | jq '.items[] | {id, name, status}'
# All nodes should show status: "RUNNING"5. One skill per domain, not one skill per command. We tried granular skills early on — one for peers, one for orderers, one for networks. It fractured the workflow. Consolidating into chaindeploy-cli (full CLI reference) and deploy-chainlaunch (full deployment guide) worked far better.
6. Version-control skills with the code. Skills live in .claude/skills/ inside the repo. When CLI flags change in a PR, the skill updates in the same PR. They never drift.
How Do You Try This Right Now?
You don't need blockchain experience. You don't need to install Go or configure Docker manually. Here's the fastest path from zero to a running blockchain network managed by Claude Code skills.
Step 1: Provision a Server (One Command)
SSH into a fresh Ubuntu VPS (Hetzner CX21, AWS t3.medium, DigitalOcean — anything with 2+ vCPU and 4GB RAM) and run:
curl -fsSL https://chainlaunch.dev/deploy.sh | bashThis interactive installer handles everything: Docker installation, ChainLaunch binary download, admin credentials, TLS configuration, and service startup. You'll have ChainLaunch running in under 3 minutes.
The script supports both Community and Pro editions, auto-detects your platform (macOS and Linux), and generates encrypted admin credentials. It works with systemd on Linux or runs in foreground mode on macOS for local development.
Step 2: Install Claude Code
If you don't have Claude Code yet:
npm install -g @anthropic-ai/claude-codeStep 3: Clone and Enter the Project
git clone https://github.com/LF-Decentralized-Trust-labs/chaindeploy
cd chaindeployThe .claude/skills/ directory is already populated with all six skills. Claude Code loads them automatically when you start a session in this directory.
Step 4: Ask Claude Code to Build Something
Start Claude Code and make a request in plain English:
claude
> Create a 4-node Besu testnet called "supply-chain" with pre-funded accountsClaude Code reads the chaindeploy-cli skill, finds the testnet besu command, and executes:
chainlaunch testnet besu --name supply-chain --nodes 4 \
--prefix besu-sc --mode docker \
--initial-balance "0xf39Fd6e51aad88F6F4ce6aB8827279cffFb92266=0x3635C9ADC5DEA00000"It then verifies all nodes are running and blocks are being produced. You've got a working QBFT network.
Want Fabric instead? Same workflow:
> Create a Fabric network with org PharmaCorp, 2 peers, 3 orderers,
and deploy a drug tracking chaincodeClaude Code handles org creation, node provisioning, channel setup, anchor peers, connection profile generation, chaincode writing, and deployment — all from the skills.
For step-by-step walkthroughs of each blockchain type, see our Besu PoC tutorial and Fabric PoC tutorial.
What Does the Agent Architecture Look Like?
Skills are one half of ChainLaunch's Claude Code integration. The other half is a roster of 12 specialized agents, each with defined roles and access boundaries.
| Agent | What It Does | Can Write Code? |
|---|---|---|
implementer-go |
Writes Go backend features | Yes |
test-engineer |
Writes unit, integration, and E2E tests | Yes (test files only) |
security-auditor |
Reviews code for OWASP Top 10, auth bypass, secret exposure | No — read-only gate |
architect |
Produces Architecture Decision Records | No — design docs only |
context-scout |
Explores codebase, traces dependencies | No — read-only |
researcher |
Web research on blockchain tech and competitors | No |
The separation matters. A security-auditor that can't edit code can't be pressured into "just fixing" a vulnerability in a way that introduces a new one. An architect that only produces ADRs forces design decisions to be documented before implementation begins.
Skills tell agents what to do. Agent definitions tell them what they're allowed to do. The combination gives you an AI-powered development workflow with actual guardrails.
How Can You Build Skills for Your Own Project?
You don't need a blockchain platform to benefit from Claude Code skills. Any project with repeatable workflows — deployment procedures, review checklists, release processes — can encode that knowledge into skills.
Start Here
Create the directory structure:
mkdir -p .claude/skills/my-first-skillWrite a SKILL.md:
---
name: my-first-skill
description: >
Describe when this skill should activate.
Be specific about the workflow it covers.
---
# Skill Title
## When to Use
- Trigger condition 1
- Trigger condition 2
## Steps
### 1. First Action
Exact command with real flags and values.
### 2. Verify
How to confirm the action worked.
### 3. Handle Failures
What to do when things go wrong.Commit it to your repo. Claude Code picks it up automatically in the next session.
Skill Design Checklist
Before shipping a skill, verify:
- Every command uses concrete examples, not placeholders
- The full workflow is documented end to end
- Verification steps follow every action
- API-only operations are called out explicitly
- Edge cases and failure modes are covered
- The skill is in version control alongside the code it documents
What's the Business Case for Skills?
Gartner predicts 90% of enterprise software engineers will use AI code assistants by 2028, up from less than 14% in early 2024 (Gartner, 2025). But here's the catch: a Gartner survey found that 42% of engineering staff report AI productivity gains of only 1-10%, and less than half of purchased AI coding assistant licenses see active use after several months.
Why do most AI coding tools underdeliver? Because they lack project-specific context. A generic AI assistant doesn't know your CLI flags, your deployment steps, or the six things that have to happen in order after creating a Fabric network. Skills fix that.
For a project manager evaluating blockchain infrastructure, the question isn't "is this technically possible?" — it's "can my team actually operate this?" Skills shift the answer from "only if we hire a blockchain DevOps specialist" to "anyone with Claude Code access."
Here's what changes concretely:
Onboarding time drops. A new team member doesn't need to read 50 pages of Fabric documentation. They ask Claude Code to "create a testnet," and the skill handles the complexity. The learning happens through doing, not reading.
Operational knowledge stops being tribal. When your senior engineer documents the Fabric lifecycle in a skill, that knowledge is available to every team member through Claude Code. It doesn't disappear when someone goes on vacation or leaves the company.
PoC velocity increases. The gap between "let's test this blockchain idea" and "here's a working demo for stakeholders" shrinks from weeks to hours. More ideas get tested. Better decisions get made.
Consistency improves. Every deployment follows the same steps. Every code review checks the same criteria. Every changelog follows the same format. Skills eliminate the variance that comes from different people remembering different subsets of the process.
Frequently Asked Questions
Do I need blockchain experience to use ChainLaunch with Claude Code skills?
No. The skills encode the operational expertise needed to deploy and manage Fabric and Besu networks. You describe what you want in plain English, and Claude Code translates that into the correct CLI commands, API calls, and verification steps. The deploy script handles server provisioning with zero blockchain knowledge required.
Can I create custom skills for my own project's workflows?
Yes. Skills are markdown files in .claude/skills/<name>/SKILL.md with YAML frontmatter. Any repeatable workflow — deployment, review, release, debugging — can be encoded as a skill. Claude Code loads them automatically when you start a session in the project directory.
How do skills differ from CLAUDE.md project instructions?
CLAUDE.md files set global project context — architecture decisions, coding conventions, repo structure. Skills are workflow-specific instruction sets invoked for particular tasks. Think of CLAUDE.md as "what this project is" and skills as "how to do specific things in this project."
What server specs do I need to run ChainLaunch?
A VPS with 2+ vCPUs and 4GB+ RAM handles most PoC workloads. Hetzner CX21 ($5/month), AWS t3.medium, or DigitalOcean droplets all work. The deploy.sh script supports Ubuntu 24.04 LTS and handles Docker, binary installation, and service configuration automatically.
Can skills call other skills or chain together?
Skills themselves don't call each other, but agents can. ChainLaunch's technical-docs skill orchestrates four sub-agents that each handle different documentation types. The skill defines the workflow; the agents execute each step with their own specialized capabilities and access boundaries.
What Comes Next
Claude Code skills turn operational expertise into something reusable, version-controlled, and accessible to every team member. For blockchain infrastructure — where the operational bar is notoriously high — this changes who can participate in PoC development and network management.
If you want to try it yourself:
- Provision ChainLaunch in one command:
curl -fsSL https://chainlaunch.dev/deploy.sh | bash - Install Claude Code:
npm install -g @anthropic-ai/claude-code - Clone the repo:
git clone https://github.com/LF-Decentralized-Trust-labs/chaindeploy && cd chaindeploy - Start building: Ask Claude Code to create a testnet, deploy chaincode, or review a PR
The skills are open source. Read them, fork them, adapt the patterns for your own projects. The best skill is the one that saves your team from repeating the same explanation for the twentieth time.
For deeper dives into the blockchain side, explore our Fabric vs Besu comparison, QBFT consensus guide, or supply chain blockchain selection guide.