Building a good AI setup takes real work. harness-kit makes it portable. Package your plugins, skills, MCP servers, hooks, and conventions into a config you can apply to any harness on any machine — and share with your team in one file. Works with Claude Code today. Designed to travel.
/plugin marketplace add harnessprotocol/harness-kit
Fallback: install with script (skills only)
If your Claude Code build doesn't support the plugin marketplace:
curl -fsSL https://raw.githubusercontent.com/harnessprotocol/harness-kit/main/install.sh | bashDownloads skill files to ~/.claude/skills/ over HTTPS. This installs only the skill files; the full plugin (e.g. index rebuild script) comes from the marketplace install.
Install explain — no dependencies, works in any codebase:
/plugin install explain@harness-kit
Then try it:
/explain src/auth/middleware.ts # explain a specific file
/explain the payment processing flow # search the codebase for a concept
/explain src/services/ # map a directory
Produces a structured explanation: summary, key components, how it connects, patterns, gotchas, and where to start if you need to change it.
| Plugin | What it does | Try it |
|---|---|---|
explain |
Layered explanations of files, functions, directories, or concepts | /explain src/auth/ |
review |
Code review for a branch, PR, or path — structured output with severity labels | /review |
docgen |
Generate or update README, API docs, architecture overview, or changelog | /docgen readme |
research |
Process any source into a structured, compounding knowledge base | /research https://... |
data-lineage |
Trace column-level data lineage through SQL, Kafka, Spark, and JDBC codebases | /data-lineage orders.amount |
orient ¹ |
Topic-focused session orientation across graph, knowledge, journal, and research | /orient auth |
capture-session ¹ |
Capture session information into a staging file for later reflection | /capture-session |
harness-share |
Export your plugin setup to harness.yaml, import it anywhere, and validate against the Harness Protocol v1 spec |
/harness-export · /harness-import · /harness-validate |
¹ Personal-workflow plugins designed for projects using the knowledge graph + journal pattern.
Plugins from the broader ecosystem that complement the kit. These are developed and maintained by their respective authors.
| Plugin | Author | What it does | Install |
|---|---|---|---|
superpowers |
Jesse Vincent | Structured dev workflows — TDD, systematic debugging, brainstorming-before-coding, subagent delegation, and git worktree isolation | See below |
/plugin marketplace add obra/superpowers-marketplace
/plugin install superpowers@obra
Capture your installed plugins into a harness.yaml file, commit it to your dotfiles, and restore it on any machine — or share it with a teammate.
/harness-export # write harness.yaml from your current setup
/harness-import harness.yaml # interactive wizard to pick what to install
/harness-validate # validate harness.yaml against the Harness Protocol v1 schema
The import wizard shows each plugin with its description and lets you pick a subset — your config is a starting point, not a mandate.
harness.yaml follows the Harness Protocol v1 open spec — a vendor-neutral format for portable AI coding harnesses that can declare plugins, MCP servers, environment variables, and instructions in one shareable file.
Shell fallback (no Claude Code required):
curl -fsSL https://raw.githubusercontent.com/harnessprotocol/harness-kit/main/harness-restore.sh | bash -s -- harness.yamlSee harness.yaml.example for the config format.
harness-kit targets Claude Code, but SKILL.md files are plain markdown — copy them into any tool's instruction system. VS Code Copilot reads CLAUDE.md natively via the chat.useClaudeMdFile setting, so the conventions guide works without modification. For per-tool setup (Copilot, Cursor, Windsurf, MCP), see the Cross-Harness setup guide.
A standalone guide to organizing Claude Code configuration with separation of concerns. Worth reading even if you don't use any plugins.
| File | Scope | Purpose |
|---|---|---|
CLAUDE.md |
Global + Project + Package | Operational: build, test, architecture |
AGENT.md |
Global + Project | Behavioral: tone, autonomy, workflow |
SOUL.md |
Global only | Identity: values, relationship, memory |
Read the full guide: Claude Conventions
Plugins can also include agent definitions — isolated specialist workers with their own context and scoped tools. See Understanding Agents for how agents, AGENT.md, and subagent definitions relate.
See CONTRIBUTING.md for plugin guidelines, skill conventions, and PR process.
- FAQ — What is this, why do I need it, how does it work
- Plugins vs. Skills — Why everything ships as a plugin, even when it's just a prompt
- Claude Conventions — Organizing Claude Code config with separation of concerns
- Understanding Agents — Disambiguating AGENT.md, custom subagents, and "AI agent" generally
Apache 2.0
