Help us improve
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
By rafa1off
Multi-agent orchestrator: 8 agents (reader, researcher, writer, thinker, checker, reviewer, tester, documenter), orchestrator routing skill, stack-agnostic dev-tools MCP server, and SessionStart hook. Activate with: Skill("/orchestrator-core:orchestrator") in CLAUDE.md.
npx claudepluginhub rafa1off/orchestrator --plugin orchestrator-coreUse this agent to run lint, typecheck, and build checks after code has been written. Returns a structured pass/fail report. Never invoke before writer has produced changes. <example> Context: Orchestrator needs to verify code written by the writer passes all checks. user: [orchestrator passes task description and list of modified files] assistant: [checker runs checks and returns structured pass/fail with error details] </example>
Use this agent to update docs or inline comments after a feature has been implemented, tested, and approved. Only invoke when the task changes a public API, module behavior, environment variable, or core architecture. Never touches source logic. <example> Context: A new feature was added. Orchestrator needs docs updated. user: [orchestrator passes diff and description of what changed] assistant: [documenter updates relevant doc files only] </example>
Use this agent to map relevant code paths and produce a structured context snapshot before writing or reviewing code. Invoke when you need to understand which files are involved in a task, what interfaces exist, and what conventions are in use — without making any changes. <example> Context: Orchestrator needs to understand which files to touch before implementing a new feature. user: [orchestrator passes task description] assistant: [reader returns structured snapshot of relevant files, interfaces, and conventions] </example>
Use this agent to find external patterns, library APIs, and internal prior decisions relevant to a task. Invoke when the task requires knowledge of external library APIs or when the task may have prior art in the project docs. <example> Context: Orchestrator needs to know how a library handles a specific pattern before writing code. user: [orchestrator passes task + research question] assistant: [researcher returns concise findings from web and docs] </example>
Use this agent to review code changes against project conventions after checker has passed. Returns APPROVED or a specific issue list with file paths and line references. Never auto-fixes — reports only. <example> Context: Checker passed, orchestrator needs a convention and quality review of the diff. user: [orchestrator passes diff and task context] assistant: [reviewer returns APPROVED or issue list] </example>
Execute an orchestrator-native plan from .claude/plans/, dispatching agents per the plan's agent map and enforcing the 5 invariants throughout.
Write an orchestrator-native implementation plan for a multi-step task, mapping each step to specific agents and respecting the 5 invariants. Saves to .claude/plans/.
Agent dispatch guide for this codebase. Defines the agent catalog, core invariants, and flexible working loop for any development task.
Admin access level
Server config contains admin-level keywords
Uses power tools
Uses Bash, Write, or Edit tools
Share bugs, ideas, or general feedback.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review
Reliable automation, in-depth debugging, and performance analysis in Chrome using Chrome DevTools and Puppeteer
Harness-native ECC plugin for engineering teams - 64 agents, 261 skills, 84 legacy command shims, reusable hooks, rules, MCP conventions, and operator workflows for Claude Code plus adjacent agent harnesses
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Tools to maintain and improve CLAUDE.md files - audit quality, capture session learnings, and keep project memory current.
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Python LSP via Astral ty — go-to-definition, find-references, document-symbol on Python files. Requires: uv tool install ty
TypeScript/JavaScript LSP via vtsls — go-to-definition, find-references, document-symbol on .ts/.tsx/.js files. Requires: npm install -g @vtsls/language-server
A private Claude Code plugin marketplace for the orchestrator multi-agent development ecosystem.
| Plugin | Description | Requires |
|---|---|---|
orchestrator-core | 8 agents, 3 skills, stack-agnostic dev-tools MCP, SessionStart hook | uv |
ty-lsp | Python LSP via Astral ty | uv tool install ty |
vtsls-lsp | TypeScript/JavaScript LSP via vtsls | npm install -g @vtsls/language-server |
orchestrator-core)The dev-tools MCP server is a Python process launched automatically when the plugin is active. It requires either:
| Runtime | Minimum version | Notes |
|---|---|---|
uv | any recent | Preferred — server runs via uv run; dependencies resolved automatically from the bundled pyproject.toml |
| Python | 3.11+ | Fallback — server runs via python; install dependencies manually: pip install fastmcp |
If uv is present it is used automatically. If only python is available, the server falls back to the plain-Python command set — no extra configuration needed.
/plugin marketplace add rafa1off/orchestrator
Or manually in your project's .claude/settings.json or user settings (~/.claude/settings.json):
{
"extraKnownMarketplaces": {
"orchestrator": {
"source": {
"source": "github",
"repo": "rafa1off/orchestrator"
}
}
}
}
# Core orchestrator (required)
claude plugin install orchestrator-core@orchestrator
# Python LSP — install for Python projects
claude plugin install ty-lsp@orchestrator
# TypeScript/JavaScript LSP — install for TS/JS projects
claude plugin install vtsls-lsp@orchestrator
For Go, Rust, Java, and other languages, check the official Claude Code plugin marketplace:
/plugin marketplace browse
Official LSP plugins (gopls, rust-analyzer, etc.) are maintained there and install without any additional configuration.
Add one line to your project's CLAUDE.md:
Skill("/orchestrator-core:orchestrator")
That's it. No other setup required.
A complete multi-agent development harness for Claude Code. The orchestrator session acts as a coordinator, calling specialized subagents on demand.
| Agent | Model | Role |
|---|---|---|
reader | haiku | Maps code paths, returns structured context snapshots |
researcher | sonnet | Finds external patterns, library APIs, prior project decisions |
writer | sonnet | Produces minimal, focused code changes from a context block |
thinker | sonnet | Deep reasoning, tradeoff analysis, architectural decisions |
checker | haiku | Runs lint + typecheck, writes structured findings |
reviewer | sonnet | Reviews diffs against conventions, writes structured findings |
tester | sonnet | Identifies missing tests, writes them, runs the suite |
documenter | sonnet | Updates docs/ and CLAUDE.md when public surfaces change |
| Skill | When to use |
|---|---|
/orchestrator-core:orchestrator | Load at every session start — the agent routing guide |
/orchestrator-core:orchestrator-plan | Before any multi-step task — writes a plan to .claude/plans/ |
/orchestrator-core:orchestrator-execute | After plan approval — dispatches agents and enforces the 5 invariants |
Stack-agnostic tools available to checker, reviewer, and tester:
| Tool | Description |
|---|---|
lint(files) | Scoped linter run |
typecheck() | Full-project type check |
test(pattern) | Scoped test run |
write_findings(...) | Writes structured JSON to .claude/pipeline/ for the orchestrator to read |
Stack is auto-detected at server startup from marker files:
| Stack | Detection | Commands |
|---|---|---|
| Python (uv) | uv.lock | uv run ruff check / uv run mypy . / uv run pytest -x |
| Python | pyproject.toml, requirements.txt | python -m ruff check / python -m mypy . / python -m pytest -x |
| TypeScript | tsconfig.json + package.json | npx eslint / npx tsc --noEmit / npx jest |
| JavaScript | package.json | npx eslint / — / npx jest |
| Go | go.mod | go vet ./... / go build ./... / go test ./... |
| Rust | Cargo.toml | cargo clippy / cargo check / cargo test |
| Ruby | Gemfile | bundle exec rubocop / — / bundle exec rspec |
| Java (Gradle) | build.gradle | ./gradlew checkstyleMain / ./gradlew compileJava / ./gradlew test |
| Java (Maven) | pom.xml | mvn checkstyle:check / mvn compile / mvn test |
Override any command by creating .claude/dev-tools.json in your project: