Explore the Docs → ·
OrchestKit Users Group →
Skill browser, demo gallery, setup wizard
Contents
Quick Start
/plugin marketplace add yonatangross/orchestkit
/plugin install ork
Then start your personalized onboarding:
/ork:setup
The setup wizard scans your codebase, detects your tech stack, recommends skills for your needs, configures MCP servers, and creates a readiness score — all in one command.
Why OrchestKit?
Every Claude Code session starts from zero. You explain your stack, patterns, preferences—again and again.
OrchestKit gives Claude persistent knowledge of production patterns that work automatically:
| Without | With OrchestKit |
|---|
| "Use FastAPI with async SQLAlchemy 2.0..." | "Create an API endpoint" → Done right |
| "Remember cursor pagination, not offset..." | Agents know your patterns |
| "Don't commit to main branch..." | Hooks block bad commits |
| "Run tests before committing..." | /ork:commit runs tests for you |
What You Get
One unified plugin, everything included.
| Component | Details |
|---|
| 111 Skills | RAG patterns, FastAPI, React 19, testing, security, database design, ML integration — loaded on-demand, zero overhead |
| 37 Agents | Specialized personas (backend-architect, frontend-dev, security-auditor) — route tasks to the right expert |
| 210 Hooks | Pre-commit checks, git protection, quality gates, browser safety — ship with confidence |
All available in a single /plugin install ork. Skills load on-demand. Hooks work automatically.
Browse everything in the Docs →
Key Commands
/ork:setup # Personalized onboarding wizard
/ork:implement # Full-stack implementation with parallel agents
/ork:expect # Diff-aware AI browser testing
/ork:review-pr # PR review with parallel agents
/ork:verify # Multi-agent validation
/ork:commit # Conventional commit with pre-checks
/ork:explore # Analyze unfamiliar codebase
/ork:remember # Save to persistent memory
/ork:doctor # Health check
Configuration
/ork:configure
Recommended MCP Servers
| Server | Purpose | Required? |
|---|
| Context7 | Up-to-date library docs | Recommended |
| Memory | Knowledge graph persistence | Recommended |
| Sequential Thinking | Structured reasoning for subagents | Optional |
| Tavily | Web search and extraction | Optional |
The setup wizard (/ork:setup) will recommend MCPs based on your stack.
Install
/plugin install ork
No tiering. No version confusion. Just one powerful plugin.
Not on Claude Code? Pull the skills into any agent (Cursor, Codex, OpenCode, …) via skills.sh:
npx skills add yonatangross/orchestkit
Release Channels
| Channel | Stability | Install |
|---|
| Stable | Production-ready | /plugin install ork |
| Beta | May have rough edges | See below |
| Alpha | Experimental, may break | See below |
To install beta or alpha:
# Beta channel
/plugin marketplace add yonatangross/orchestkit --ref beta --name orchestkit-beta
/plugin install ork@orchestkit-beta
# Alpha channel
/plugin marketplace add yonatangross/orchestkit --ref alpha --name orchestkit-alpha
/plugin install ork@orchestkit-alpha
Run /ork:doctor to check which channel you're on. Full docs.
FAQ
Plugin not found?
/plugin list
/plugin uninstall ork && /plugin install ork
Hooks not firing?
Run /ork:doctor to diagnose.
Claude Code version?
Requires ≥2.1.148 (supported floor; Opus 4.8, xhigh effort, dynamic workflows, sandbox.network.deniedDomains, native binary, hardened Bash(rm:*)/Bash(find:*) rules). Check with claude --version.
Development
npm run build # Build plugins from src/
npm test # Run all tests
Edit src/ and manifests/, never plugins/ (generated).
See CONTRIBUTING.md for details.
What's New
v8.16.0 — CC 2.1.149 → 2.1.161 adoption catch-up
- CC 2.1.160 + 2.1.161 adopted — shell/build-config write prompts, single-file
grep satisfies read-before-edit, background-session SIGTERM-before-SIGKILL, dynamic-workflow trigger renamed workflow → ultracode, parallel-tool independent failure, claude mcp secret redaction, OTEL_RESOURCE_ATTRIBUTES metric labels, claude agents done/total, /mcp collapse-unused-connectors.
- CC 2.1.157 adopted —
.claude/skills plugin auto-load (no marketplace), claude plugin init, EnterWorktree mid-session switching, worktrees unlocked on finish, claude agents --agent, backspace-dismiss workflow trigger.
- Skill tool-signature fixes — corrected
SendMessage / PushNotification examples that had drifted from the live tool schemas across 5 skills, surfaced by a 61-skill adversarially-verified staleness audit.
- Version matrix now 472 entries; supported floor held at 2.1.148 (frozen until 2026-08-23).
v7.63.0 — CC 2.1.116 Adoption + gh rate-limit enforcement
MIN_CC_VERSION bumped 2.1.113 → 2.1.116, 10 new matrix entries (333 total), including agent_hooks_main_thread (agent-scoped hooks fire in --agent main thread, not just subagent), sandbox_rm_dangerous_path_fix, reload_plugins_auto_deps, bash_gh_rate_limit_hint- New PostToolUse hook
gh-rate-limit-tracker — detects primary/secondary/HTTP-403+rate-phrase signals in gh stdout, injects additionalContext telling the model to stop the loop and wait for reset
- 5 new skill evals covering the 2.1.116 behavioral knowledge in
github-operations, create-pr, review-pr, doctor, and agent-hook authoring
- Hook count 180 → 181; 14 agent hooks audited for main-thread firing safety (all defensive blockers — net positive)
v7.38.0 — CC 2.1.101 Full Adoption + Frontmatter Audit
- 26 skill frontmatter fixes — CC 2.1.101 enforces
context:fork and agent: fields that were previously silently ignored; audited and corrected all 105 skills
- 18 new CC features tracked — 253 total in version matrix (deny-overrides-ask, dynamic MCP in subagents, worktree file access, focus mode,
/team-onboarding)
- Wrong agent assignments fixed — performance→frontend-performance-engineer, devops→ci-cd-engineer
- agent: removed from commit/create-pr — prevents CC from delegating to agent instead of running skill workflow
v7.37.0 — CC 2.1.98 Full Adoption
- Monitor tool — real-time streaming from background builds/tests in implement, cover, verify, review-pr
- Partial result synthesis — crashed background agents report progress; 4 workflow skills salvage partial work instead of re-spawning
- Subprocess hardening —
SCRIPT_CAPS=500 enforced, PID namespace sandbox, ENV_SCRUB active
- 17 new CC features tracked — 235 total in version matrix
- Anti-sycophancy protocol — bans performative agreement in review skills (from Superpowers analysis)
- Verification gate — cross-cutting 5-step evidence rule across 7 workflow skills
- Agent status protocol — DONE/DONE_WITH_CONCERNS/BLOCKED/NEEDS_CONTEXT across all 39 agents
- Pressure-test framework — RED-GREEN behavioral scenarios for 5 discipline skills
v7.36.0 — CORS Hardening + Release-Please Fix
- CORS restricted —
Access-Control-Allow-Origin from * to same-origin on docs site
- Release-please drift fix —
x-release-please-version marker eliminates build drift on release PRs
v7.33.0 — CC 2.1.97 Full Utilization
- 26 new CC features tracked — refreshInterval status line, workspace.git_worktree, TRACEPARENT OTEL, Stop/SubagentStop long-session fix, and 20 more
v7.30.0 — CC 2.1.94 Integration + Skill Hook Unlock
- 20 dead skill frontmatter hooks activated — CC 2.1.94 fixed silent ignore; context loaders now fire across 15 workflow skills
- Dynamic session titles — prompt bar shows
{branch} · {effort} via hookSpecificOutput.sessionTitle
- Effort default aligned —
high default matches CC 2.1.94+ for API-key/Bedrock/Vertex/Team/Enterprise users
- 105 skills, 39 agents, 1 hooks, 25 invocable commands
v7.27.0 — shadcn/ui v4 Style System + CC 2.1.90
- shadcn/ui v4 styles — 6 styles across 7 design skills
- CC 2.1.90 integration — format-on-save, offline resilience,
/powerup onboarding
Full Changelog →
Community
Join the Code with Yonatan community for AI dev tips, OrchestKit support, and connecting with other builders:
