Skip to content
Back to search

orchestkit

MarketplaceWorks with Claude CodeGitHub

The Complete AI Development Toolkit - 115 skills, 36 agents, 47 commands, 163 hooks for full-stack development

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

Its scripts were only partly scanned.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/yonatangross/orchestkit with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/yonatangross/orchestkit'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/yonatangross-orchestkit/badge.svg)](/listings/yonatangross-orchestkit)

HTML

<a href="/listings/yonatangross-orchestkit"><img src="/api/listings/yonatangross-orchestkit/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Pick the host you actually use. Claude Code is the full plugin (skills + agents + hooks). Cursor gets the same `ork` plugin minus Claude hook scripts. skills.sh is skills only — start with the 12 below, not the whole catalog.

Host support matrix

Measured 2026-09-08 on pi 0.85, Codex CLI and cursor-agent. Details, commands and the lane model: [OrchestKit on pi, Codex and Cursor](https://orchestkit.dev/docs/guides/orchestkit-on-pi-codex-cursor). Antigravity measured 2026-09-18 on agy 1.2.6; full evidence in `docs/audits/agy-host-support-2026-09-18.md`.

| Host | Skills | Agents | Hooks | Install | Needs | |---|---|---|---|---|---| | Claude Code | 115 | 36 | 163 | plugin (CLI) | Claude Code 2.1.277 or newer | | Cursor | 115 | 36 | not yet (#4004) | plugin (Settings) | | | Codex | 9 | 4 roles | not yet (#4004) | ork-codex pack | | | Devin | 82 | ? | ? | plugin (git) | | | OpenCode | 12 | ? | ? | skills.sh | | | Muse Code | 12 | ? | ? | .agents/skills | | | Pi | 82 | ? | not yet (#4004) | pi manifest | | | Antigravity | 12 | ? | ? | .agents/skills | |

`?` means not measured yet, never zero. `not yet` means a tracked issue will bring it.

How each host wires the rest (measured, see the dates above):

| Surface | Claude Code | Cursor | Codex | pi | Devin | Antigravity | |---|---|---|---|---|---|---| | Rules | repo convention | 14, plugin `rules` key | `AGENTS.md` | none | `AGENTS.md`, always on | `AGENTS.md`, per-directory | | Commands | `/ork:<skill>` | 36 wrappers | `$ork-<skill>` | `/skill:<name>` | `/ork:<skill>` | `/<skill-name>`; `ork:<skill>` via plugin install | | MCP config | `.mcp.json` | `.cursor/mcp.json` | plugin `mcp.json` | `.pi/mcp.json` | `mcp.json` / `.mcp.json` | `~/.gemini/config/mcp_config.json` or plugin `mcp_config.json`; repo `.mcp.json` not read | | Status | shipped | shipped | shipped | shipped | skills only, GH-4146 | skills only, measured agy 1.2.6 |

Claude Code

/plugin marketplace add yonatangross/orchestkit
/plugin install ork

Then `/ork:setup`. The wizard scans the repo, recommends skills, and writes MCP config.

CLI equivalent: `claude plugin marketplace add yonatangross/orchestkit && claude plugin install ork@orchestkit`.

Channels: `ork` is the stable channel, pinned to the latest release. `ork-alpha` is the same plugin tracking `main`, in the same marketplace (`claude plugin install ork-alpha@orchestkit`); use it only if you want changes before they are released.

Cursor

Settings → Plugins / marketplaces → add `yonatangross/orchestkit` → enable **ork** → **open a new chat**. Same plugin Claude Code installs, not a five-skill fork. See [Install → Cursor](#cursor).

skills.sh (Cursor, Codex, OpenCode, …)

Starter 12 from measured installs plus the ork skills we retain (full why table under [Install](#install)):

npx skills add yonatangross/orchestkit -s devops-deployment -s responsive-patterns -s architecture-decision-record -s ui-components -s rag-retrieval -s agent-orchestration -s brainstorm -s expect -s auto -s review-pr -s fix-issue -s commit

Full catalog: `npx skills add yonatangross/orchestkit` (hundreds of SKILL.md files; do not treat that as unique users). Live install totals stay on the skills.sh badge above.

---

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 128 lines
  1. {
  2. "$schema": "https://anthropic.com/claude-code/marketplace.schema.json",
  3. "name": "orchestkit",
  4. "version": "10.14.0",
  5. "description": "The Complete AI Development Toolkit - 115 skills, 36 agents, 47 commands, 163 hooks for full-stack development",
  6. "owner": {
  7. "name": "Yonatan Gross",
  8. "email": "yonaigross@gmail.com"
  9. },
  10. "metadata": {
  11. "description": "AI-assisted development plugin with built-in best practices, security patterns, quality gates, and multi-agent orchestration for Claude Code.",
  12. "pluginRoot": "./plugins"
  13. },
  14. "plugins": [
  15. {
  16. "name": "ork",
  17. "version": "10.14.0",
  18. "description": "Complete AI development toolkit covering AI/LLM, backend, frontend, security, and testing patterns, with specialized agents, commands, and lifecycle hooks.",
  19. "author": {
  20. "name": "Yonatan Gross",
  21. "email": "yonaigross@gmail.com"
  22. },
  23. "category": "development",
  24. "source": {
  25. "source": "git-subdir",
  26. "url": "https://github.com/yonatangross/orchestkit.git",
  27. "path": "plugins/ork",
  28. "ref": "v10.14.0"
  29. }
  30. },
  31. {
  32. "name": "ork-alpha",
  33. "version": "10.14.0",
  34. "description": "Prerelease channel for OrchestKit. Tracks main, so it carries alpha and beta builds ahead of the stable ork plugin. Install only if you want unreleased changes.",
  35. "author": {
  36. "name": "Yonatan Gross",
  37. "email": "yonaigross@gmail.com"
  38. },
  39. "category": "development",
  40. "source": {
  41. "source": "git-subdir",
  42. "url": "https://github.com/yonatangross/orchestkit.git",
  43. "path": "plugins/ork",
  44. "ref": "main"
  45. }
  46. },
  47. {
  48. "name": "ork-extras",
  49. "version": "10.14.0",
  50. "description": "Opt-in reference skills split out of ork: library patterns (LangGraph, RAG, MCP, Vite, Zustand, Storybook and more) and engineering domain guides (accessibility, distributed systems, DDD, release management). Skills only, no agents or hooks. Install next to ork; invoke as /ork-extras:<skill>.",
  51. "author": {
  52. "name": "Yonatan Gross",
  53. "email": "yonaigross@gmail.com"
  54. },
  55. "category": "development",
  56. "source": {
  57. "source": "git-subdir",
  58. "url": "https://github.com/yonatangross/orchestkit.git",
  59. "path": "plugins/ork-extras",
  60. "ref": "v10.14.0"
  61. }
  62. },
  63. {
  64. "name": "promote-lights",
  65. "description": "Mod (function hooks): CI lights above the prompt while a promote PR is open, or for any PR with /lights watch owner/repo#N; /lights [off|refresh|watch]. Requires Claude Code 2.1.266+. On 2.1.266 to 2.1.286 set CLAUDE_CODE_ENABLE_FUNCTION_HOOKS=1; on 2.1.287+ Mods are public.",
  66. "version": "0.0.15",
  67. "author": {
  68. "name": "Yonatan Gross",
  69. "email": "yonaigross@gmail.com"
  70. },
  71. "category": "development",
  72. "source": {
  73. "source": "git-subdir",
  74. "url": "https://github.com/yonatangross/orchestkit.git",
  75. "path": "mods/promote-lights",
  76. "ref": "main"
  77. }
  78. },
  79. {
  80. "name": "secrets-veil",
  81. "description": "Mod (function hooks): masks secret values in tool results before the model reads them. Requires Claude Code 2.1.266+. On 2.1.266 to 2.1.286 set CLAUDE_CODE_ENABLE_FUNCTION_HOOKS=1; on 2.1.287+ Mods are public.",
  82. "version": "0.0.11",
  83. "author": {
  84. "name": "Yonatan Gross",
  85. "email": "yonaigross@gmail.com"
  86. },
  87. "category": "security",
  88. "source": {
  89. "source": "git-subdir",
  90. "url": "https://github.com/yonatangross/orchestkit.git",
  91. "path": "mods/secrets-veil",
  92. "ref": "main"
  93. }
  94. },
  95. {
  96. "name": "memory-lens",
  97. "description": "Mod (function hooks): recalls the few memories related to each prompt you send (not notifications, peers, schedules or slash commands, and never on file tool calls); /memory-lens for stats, reload and show. Requires Claude Code 2.1.266+. On 2.1.266 to 2.1.286 set CLAUDE_CODE_ENABLE_FUNCTION_HOOKS=1; on 2.1.287+ Mods are public.",
  98. "version": "0.1.1",
  99. "author": {
  100. "name": "Yonatan Gross",
  101. "email": "yonaigross@gmail.com"
  102. },
  103. "category": "development",
  104. "source": {
  105. "source": "git-subdir",
  106. "url": "https://github.com/yonatangross/orchestkit.git",
  107. "path": "mods/memory-lens",
  108. "ref": "main"
  109. }
  110. },
  111. {
  112. "name": "prompt-enhancer",
  113. "description": "Beta mod (function hooks, experimental): a band above the prompt rewrites your draft with the session haiku model into a short, stack-aware prompt (typo fixes, your real subagents, skills and MCP servers, one Done when line, no questions); you edit the rewrite, then Send or Undo. The catalog prefix is prompt-cached. Each Enhance click costs model usage on your account (1024 token cap). Requires Claude Code 2.1.300+.",
  114. "version": "0.2.0",
  115. "author": {
  116. "name": "Yonatan Gross",
  117. "email": "yonaigross@gmail.com"
  118. },
  119. "category": "development",
  120. "source": {
  121. "source": "git-subdir",
  122. "url": "https://github.com/yonatangross/orchestkit.git",
  123. "path": "mods/prompt-enhancer",
  124. "ref": "main"
  125. }
  126. }
  127. ]
  128. }
README.md · README · 797 lines
  1. <!-- markdownlint-disable MD033 MD041 -->
  2. <div align="center">
  3. <img src="docs/banner.png" alt="OrchestKit - Stop explaining your stack. Start shipping." width="100%" />
  4. **<!--ork:skills-->115<!--/ork--> skills · <!--ork:agents-->36<!--/ork--> agents · <!--ork:hooks-->163<!--/ork--> hooks**
  5. [![Claude Code](https://img.shields.io/badge/Claude_Code-≥2.1.277-7C3AED?style=for-the-badge&logo=anthropic)](https://claude.ai/claude-code)
  6. [![License](https://img.shields.io/badge/License-MIT-yellow?style=for-the-badge)](./LICENSE)
  7. [![GitHub Stars](https://img.shields.io/github/stars/yonatangross/orchestkit?style=for-the-badge&logo=github)](https://github.com/yonatangross/orchestkit)
  8. [![Community](https://img.shields.io/badge/Community-WhatsApp-25D366?style=for-the-badge&logo=whatsapp)](https://platform.yonyon.ai/circle?ref=readme)
  9. [![Ask DeepWiki](https://img.shields.io/badge/Ask-DeepWiki-1A1A2E?style=for-the-badge&logo=bookstack&logoColor=4F9CF9)](https://deepwiki.com/yonatangross/orchestkit)
  10. [![MCP Toplist](https://mcptoplist.com/badge/io.github.yonatangross%2Forchestkit.svg)](https://mcptoplist.com/server/io.github.yonatangross%2Forchestkit)
  11. [![skills.sh](https://www.skills.sh/b/yonatangross/orchestkit)](https://www.skills.sh/yonatangross/orchestkit)
  12. [![Vercel OSS Program](https://vercel.com/oss/program-badge-2026.svg)](https://vercel.com/open-source-program)
  13. </div>
  14. ## Start here
  15. OrchestKit is free and open source (MIT, no account): skills and agents for Claude Code, Cursor, Codex and five more coding agents, plus hooks in Claude Code, for review, testing, security and memory. [Where it runs](https://orchestkit.dev/docs/compatibility).
  16. ```bash
  17. claude plugin marketplace add yonatangross/orchestkit && claude plugin install ork@orchestkit
  18. ```
  19. [Docs](https://orchestkit.dev/docs) · [What's new in v10](https://orchestkit.dev/docs/whats-new-v10) · [yonyon.ai](https://yonyon.ai)
  20. <!-- X account link: add it here when the account exists. -->
  21. ---
  22. <p align="center">
  23. <a href="https://orchestkit.dev/"><strong>Explore the Docs →</strong></a> ·
  24. <a href="https://yonyon.ai/go/orchestkit?utm_campaign=readme"><strong>OrchestKit Community →</strong></a><br>
  25. <sub>Skill browser, demo gallery, setup wizard</sub>
  26. </p>
  27. ---
  28. ## Contents
  29. - [Quick Start](#quick-start)
  30. - [Why OrchestKit?](#why-orchestkit)
  31. - [What You Get](#what-you-get)
  32. - [Key Commands](#key-commands)
  33. - [Configuration](#configuration)
  34. - [What OrchestKit observes](#what-orchestkit-observes)
  35. - [Install](#install)
  36. - [FAQ](#faq)
  37. - [Development](#development)
  38. - [What's New](#whats-new)
  39. - [Community](#community)
  40. ## Quick Start
  41. Pick the host you actually use. Claude Code is the full plugin (skills + agents + hooks). Cursor gets the same `ork` plugin minus Claude hook scripts. skills.sh is skills only — start with the 12 below, not the whole catalog.
  42. #### Host support matrix
  43. Measured 2026-09-08 on pi 0.85, Codex CLI and cursor-agent. Details, commands and the lane model: [OrchestKit on pi, Codex and Cursor](https://orchestkit.dev/docs/guides/orchestkit-on-pi-codex-cursor). Antigravity measured 2026-09-18 on agy 1.2.6; full evidence in `docs/audits/agy-host-support-2026-09-18.md`.
  44. <!-- compat-table:start (generated by scripts/stamp-compat-table.js) -->
  45. | Host | Skills | Agents | Hooks | Install | Needs |
  46. |---|---|---|---|---|---|
  47. | Claude Code | 115 | 36 | 163 | plugin (CLI) | Claude Code 2.1.277 or newer |
  48. | Cursor | 115 | 36 | not yet (#4004) | plugin (Settings) | |
  49. | Codex | 9 | 4 roles | not yet (#4004) | ork-codex pack | |
  50. | Devin | 82 | ? | ? | plugin (git) | |
  51. | OpenCode | 12 | ? | ? | skills.sh | |
  52. | Muse Code | 12 | ? | ? | .agents/skills | |
  53. | Pi | 82 | ? | not yet (#4004) | pi manifest | |
  54. | Antigravity | 12 | ? | ? | .agents/skills | |
  55. `?` means not measured yet, never zero. `not yet` means a tracked issue will bring it.
  56. <!-- compat-table:end -->
  57. How each host wires the rest (measured, see the dates above):
  58. | Surface | Claude Code | Cursor | Codex | pi | Devin | Antigravity |
  59. |---|---|---|---|---|---|---|
  60. | Rules | repo convention | 14, plugin `rules` key | `AGENTS.md` | none | `AGENTS.md`, always on | `AGENTS.md`, per-directory |
  61. | Commands | `/ork:<skill>` | 36 wrappers | `$ork-<skill>` | `/skill:<name>` | `/ork:<skill>` | `/<skill-name>`; `ork:<skill>` via plugin install |
  62. | MCP config | `.mcp.json` | `.cursor/mcp.json` | plugin `mcp.json` | `.pi/mcp.json` | `mcp.json` / `.mcp.json` | `~/.gemini/config/mcp_config.json` or plugin `mcp_config.json`; repo `.mcp.json` not read |
  63. | Status | shipped | shipped | shipped | shipped | skills only, GH-4146 | skills only, measured agy 1.2.6 |
  64. ### Claude Code
  65. ```bash
  66. /plugin marketplace add yonatangross/orchestkit
  67. /plugin install ork
  68. ```
  69. Then `/ork:setup`. The wizard scans the repo, recommends skills, and writes MCP config.
  70. CLI equivalent: `claude plugin marketplace add yonatangross/orchestkit && claude plugin install ork@orchestkit`.
  71. Channels: `ork` is the stable channel, pinned to the latest release. `ork-alpha` is the same plugin tracking `main`, in the same marketplace (`claude plugin install ork-alpha@orchestkit`); use it only if you want changes before they are released.
  72. ### Cursor
  73. Settings → Plugins / marketplaces → add `yonatangross/orchestkit` → enable **ork** → **open a new chat**. Same plugin Claude Code installs, not a five-skill fork. See [Install → Cursor](#cursor).
  74. ### skills.sh (Cursor, Codex, OpenCode, …)
  75. Starter 12 from measured installs plus the ork skills we retain (full why table under [Install](#install)):
  76. ```bash
  77. npx skills add yonatangross/orchestkit -s devops-deployment -s responsive-patterns -s architecture-decision-record -s ui-components -s rag-retrieval -s agent-orchestration -s brainstorm -s expect -s auto -s review-pr -s fix-issue -s commit
  78. ```
  79. Full catalog: `npx skills add yonatangross/orchestkit` (hundreds of SKILL.md files; do not treat that as unique users). Live install totals stay on the skills.sh badge above.
  80. ---
  81. ## Why OrchestKit?
  82. Every Claude Code session starts from zero. You explain your stack, patterns, preferences—again and again.
  83. OrchestKit gives Claude **persistent knowledge** of production patterns that work automatically:
  84. | Without | With OrchestKit |
  85. |---------|-----------------|
  86. | "Use FastAPI with async SQLAlchemy 2.0..." | "Create an API endpoint" → Done right |
  87. | "Remember cursor pagination, not offset..." | Agents know your patterns |
  88. | "Don't commit to main branch..." | Hooks block bad commits |
  89. | "Run tests before committing..." | `/ork:commit` runs tests for you |
  90. ---
  91. ## What You Get
  92. **One unified plugin, everything included.**
  93. | Component | Details |
  94. |-----------|---------|
  95. | **<!--ork:skills-->115<!--/ork--> Skills** | RAG patterns, FastAPI, React 19, testing, security, database design, ML integration — loaded on-demand, zero overhead |
  96. | **<!--ork:agents-->36<!--/ork--> Agents** | Specialized personas (backend-architect, frontend-dev, security-auditor) — route tasks to the right expert |
  97. | **<!--ork:hooks-->163<!--/ork--> Hooks** | Pre-commit checks, git protection, quality gates, browser safety — ship with confidence |
  98. All available in a single `/plugin install ork`. Skills load on-demand. Hooks work automatically.
  99. **[Browse everything in the Docs →](https://orchestkit.dev/docs/skills/overview)**
  100. ---
  101. ## Key Commands
  102. ```bash
  103. /ork:auto # Front door: describe a goal, it routes to the right skill
  104. /ork:setup # Personalized onboarding wizard
  105. /ork:implement # Full-stack implementation with parallel agents
  106. /ork:expect # Diff-aware AI browser testing
  107. /ork:review-pr # PR review with parallel agents
  108. /ork:verify # Multi-agent validation
  109. /ork:commit # Conventional commit with pre-checks
  110. /ork:explore # Analyze unfamiliar codebase
  111. /ork:remember # Save to persistent memory
  112. /ork:doctor # Health check
  113. ```
  114. ---
  115. ## Configuration
  116. `/ork:setup` detects your stack, recommends MCP servers, and writes the configuration for you.
  117. ### Recommended MCP Servers
  118. | Server | Purpose | Required? |
  119. |--------|---------|-----------|
  120. | Context7 | Up-to-date library docs | **Prerequisite** (22 of 36 agents grant its tools) |
  121. | Memory | Knowledge graph persistence | Recommended |
  122. | Sequential Thinking | Structured reasoning for subagents | Recommended |
  123. | Tavily | Web search and extraction | Optional |
  124. Set `"alwaysLoad": true` on the first three in your `.mcp.json`. It skips the per-skill tool probe and shaves ~150ms off cold starts.
  125. **Context7 is a prerequisite, and ork does not ship it.** 22 agents grant
  126. `mcp__context7__*` in their frontmatter, but `.mcp.json` is user-owned and project-scoped,
  127. so the grant refers to a server you add. Skip it and those agents answer from training
  128. data with no error raised. The recommended entry is the hosted HTTP server, which costs
  129. no local process:
  130. ```json
  131. "context7": {
  132. "type": "http",
  133. "url": "https://mcp.context7.com/mcp"
  134. }
  135. ```
  136. Free tier: 1,000 requests, public repos, no account. Context7 Pro ($10 per seat per
  137. month) raises that to 5,000 per seat and parses private repos; add
  138. `"headers": { "Authorization": "Bearer ${CONTEXT7_API_KEY}" }` and export the `ctx7sk-`
  139. key. Add the header only once the variable is exported: with it unset the unexpanded
  140. literal is sent as the token and every query fails, and it does **not** fall back to the
  141. anonymous free tier, so the keyless entry above is strictly better than a header with no
  142. key behind it. The legacy stdio transport (`npx -y @upstash/context7-mcp@4.0.2`) is the fallback
  143. when the hosted endpoint is unreachable, but it spawns one child process per Claude Code
  144. session, so the fan-out scales with how many sessions you keep open.
  145. **Optional companion for web-platform questions:** Chrome's [Modern Web Guidance](https://github.com/GoogleChrome/modern-web-guidance) (Apache-2.0) is a separate tool that checks a site against current web-platform guidance; pin a version when you run it, for example `npx modern-web-guidance@1.0.0`.
  146. ### Customizing skills
  147. Skills install as files on your disk, but **don't hand-edit the installed copy** — it gets overwritten on update and silently diverges from the canonical playbook. The supported ways to extend (user-level skills, project skills, upstream PRs, or disabling a bundled skill) are in [docs/extending-skills.md](docs/extending-skills.md).
  148. ---
  149. ## What OrchestKit observes
  150. OrchestKit is a quality-gate plugin, so its hooks are the product rather than an
  151. add-on. This section states plainly what they see, where it goes, and how to turn
  152. each piece off.
  153. **Scope: broad and intentional.** OrchestKit registers <!--ork:hooks-->163<!--/ork--> hooks across <!--ork:events-->32<!--/ork-->
  154. lifecycle events, including `SessionStart`, `UserPromptSubmit`, `PreToolUse`,
  155. `PostToolUse`, and `Stop`. They are **not** gated to a particular framework or
  156. project type, because the gates they enforce (secret-write blocking, protected-file
  157. guards, git safety, file-size limits, agent status protocol) apply to any codebase.
  158. If you only want gates on some projects, enable the plugin per-project rather than
  159. globally.
  160. **Where data goes: a local file on your own disk.**
  161. | What | Destination | Notes |
  162. |---|---|---|
  163. | Lifecycle events (session end, PR merged, goal converged, chain phase) | `~/.local/state/orchestkit/events.jsonl` | Written unconditionally, rotated at 10 MB. `ORK_EVENTS_LOG` redirects the path (used by the test suite) |
  164. | Hook metrics: event name, tool name, payload size, duration | same local file | Size-capped metrics only |
  165. | Hook telemetry log: event, tool name, session id, branch, timestamp, and tool inputs and outputs | `telemetry/events.jsonl` in the plugin data folder (or `.claude/telemetry/` in the project) | Secrets redacted, every string cut to 500 characters. Rotated at 10 MB at session end; rotated files deleted after 7 days |
  166. | Prompt text | **Not recorded** in either log | Hooks read it to make an allow/deny decision, then discard it |
  167. | Remote sync | **Off** | No endpoint is compiled in; see below |
  168. There is deliberately **no global kill switch** for the local write, because the
  169. gates depend on that state (the git-safety and chain-staleness hooks read their
  170. own prior events). To stop it entirely, disable the plugin. Individual noisy hooks
  171. have their own opt-outs: `ORK_DISABLE_DEBT_TRACKER`, `ORK_DISABLE_WORKTREE_VERIFIER`,
  172. `ORK_DISABLE_COORDINATION_METRICS`, `ORK_NO_NOTIFY`, `ORK_NO_STALE_SWEEP`, and
  173. `ORCHESTKIT_SKIP_SLOW_HOOKS` among others.
  174. **Network access from the hooks themselves is opt-in and unset by default.** The
  175. only remote host compiled into the shipped hook bundles is `api.typesafe.ai`,
  176. used solely by the two opt-in TypeSafe settings below (`grep -o 'https\?://[A-Za-z0-9.-]*' plugins/ork/hooks/dist/*.mjs`
  177. also lists `github.com`, which appears only in printed links). By default some
  178. hooks run CLIs you already have, which use your own credentials and network:
  179. - `claude -p --model haiku` on the first prompt of a session, to title and color
  180. it. It sends the git branch name and the first 600 characters of that prompt
  181. to Anthropic through your own Claude Code login, with no tools, MCP servers or
  182. hooks in the child. `ORK_SESSION_IDENTITY=0` turns it off.
  183. - `gh issue view <n>` when a subagent starts on a branch named like
  184. `issue/<n>-...` or `fix/<n>-...`, to add the issue title and body as context.
  185. - `gh pr view` and `gh pr list` when a Bash command runs `gh pr merge` with
  186. `--delete-branch`, to warn before deleting a branch that open PRs are based on.
  187. Any other outbound call happens only if you configure a destination yourself,
  188. via one of:
  189. - `ORCHESTKIT_HOOK_URL` + `ORCHESTKIT_HOOK_TOKEN` (or the plugin's `webhookUrl` and
  190. `hookToken` options), which enable `hooks/bin/telemetry-sync.mjs`. It POSTs
  191. your local JSONL to *your own* endpoint. A `SessionEnd` hook runs it in the
  192. background; without a URL and token it exits at once and sends nothing. The
  193. same two settings enable the webhook forwarder, which streams hook events,
  194. HMAC signed, to that endpoint.
  195. - `telemetry.sinks` entries in `~/.claude/settings.local.json` or in the plugin
  196. data folder's `settings.local.json` (user scope only, never read from a
  197. project), each an `https:` URL you choose.
  198. - `ORK_HQ_TELEMETRY_URL`, which points the telemetry HTTP sink at *your own* collector.
  199. - `ORK_HQ_TELEMETRY_USE_HQ_API=1` together with `HQ_API_URL`, the same sink aimed
  200. at a self-hosted HQ API.
  201. - `ORK_SESSION_CATEGORY_PROVIDER=jev` (or `shadow`) together with the TypeSafe
  202. key variable `ORK_TYPESAFE_API_KEY`, a second classifier for the session work
  203. category. The session-identity hook already asks a local `claude -p --model haiku`
  204. process for a title and a category; with both variables set it also asks
  205. TypeSafe's Jev model (`api.typesafe.ai`, model pinned to `jev-1.13.0`) one typed
  206. Choice over the same eight categories and the same criteria text. This is the
  207. one exception to the "no hardcoded host" note above, and it is dormant unless
  208. both variables are set. What leaves your machine: the git branch name and the
  209. first 600 characters of the session's first prompt, sent to a third-party
  210. processor under its own data policy. What changes in `jev` mode: when Jev
  211. answers at confidence 0.8 or above, its category decides the session color
  212. (held out on 150 sessions, that band is 93.5% correct);
  213. below 0.8, or on any error, haiku's category decides as before. The title and
  214. emoji always come from haiku. In `shadow` mode nothing you see changes; Jev is
  215. only logged beside haiku. To turn it on locally, in the shell that launches
  216. `claude`: `export ORK_TYPESAFE_API_KEY="$(<your secret manager> ...)"` and
  217. `export ORK_SESSION_CATEGORY_PROVIDER=jev`. Both outcomes are logged once per
  218. session in the hook log, as
  219. `category jev: haiku=<cat> jev=<cat> agree=<bool> confidence=<n> decided_by=jev|haiku threshold=0.8 latency_ms=<n>`,
  220. and as `session-identity.shadow.json` with the same fields (labels, decision,
  221. timing; never prompt text) next to the raw answer `session-identity.jev.json`
  222. in the session data directory. Cost of opting in: the first prompt of a session
  223. waits for the call to settle, 1.1 to 1.7 s measured from a fresh hook process
  224. (the eval's 320 ms was a warm connection), 3 s at most before it gives up.
  225. - `ORK_ROUTE_JEV=shadow` (or `steer`) together with `ORK_TYPESAFE_API_KEY`, the
  226. Jev routing seam for `/ork:auto` (#4233). Off by default. When set, every
  227. build-shaped prompt (the same test the once-per-session executor reminder
  228. uses: an imperative build or fix verb, 40 characters or more, no explicit
  229. `/plugin:skill`) is sent to the same TypeSafe endpoint and model as one typed
  230. Choice over nineteen route classes plus four side judgments (needs a worktree,
  231. needs a browser, mutation risk, needs the operator). What leaves your machine:
  232. the first 1,500 characters of the prompt after a redactor has replaced
  233. secrets, emails, Israeli phone numbers, nine digit ids, `op://` references and
  234. the names of directories under `clients/` with `[SECRET]`, `[EMAIL]`,
  235. `[PHONE]` and `[CLIENT]`, plus the repository basename. The serialized request
  236. is scanned again before it leaves; anything that survives redaction refuses
  237. the call. In `shadow` mode nothing you see changes; the verdict is logged. In
  238. `steer` mode, at confidence `ORK_ROUTE_JEV_FLOOR` (default 0.5) or above, the
  239. one-line executor reminder names the executor the class maps to
  240. (`route: dev_fix -> /ork:fix-issue (conf 0.83)`); the model still reads the
  241. `/ork:auto` table and says whether it agrees. Fallbacks are fail open: no key,
  242. timeout, non-2xx, malformed answer, below the floor, or the daily token budget
  243. (`ORK_ROUTE_JEV_DAILY_TOKENS`, default 2,000,000) spent all mean the existing
  244. path runs untouched; a 402 or 429 switches the seam off for 24 hours. One log
  245. line per prompt,
  246. `route jev: intent=<class> conf=<n> top3=<a:p,b:p,c:p> worktree=<n> browser=<n> mutation=<n> operator=<n> floor=<n> decided_by=jev|table|off|budget|egress latency_ms=<n> input_tokens=<n> redacted=<n>`,
  247. and one record per prompt in `jev-route.jsonl` next to
  248. `session-identity.jev.json` (labels, timings and a sha256 of the redacted
  249. text; never prompt text). Offline replay: `node scripts/eval/route-check.mjs --jev`
  250. and `node scripts/eval/jev-route-score.mjs`. The vendor's agent skill is a
  251. peer plugin installed from its own marketplace, never a copied directory:
  252. `claude plugin marketplace add typesafe-ai/skills` then
  253. `claude plugin install typesafe@typesafe-ai`; `/ork:doctor` reports the
  254. installed version against the marketplace and prints the update commands.
  255. The sink returns early when the URL or the token is missing, and
  256. `telemetry-sync.mjs` prints `No ORCHESTKIT_HOOK_URL or TOKEN configured. Nothing
  257. to sync.` then exits 0. There is no analytics ping, no crash reporter, and no
  258. feature-flag fetch.
  259. **What OrchestKit never reads.** No OS keychain lookups, no `~/.aws/credentials`,
  260. no SSH private keys, no browser cookie or login stores, no clipboard. The one
  261. place secret-shaped paths appear in the source is
  262. `plugins/ork/hooks/dist/pretool.mjs`, where `id_rsa`, `.pem`, `.env`, and
  263. `credentials.json` form a **blocklist** that stops Claude writing to them. That
  264. code denies access; it does not read those files.
  265. **Third-party MCP servers are recommendations, not bundled dependencies.** The
  266. plugin ships no `.mcp.json` and declares no `mcpServers`. The table under
  267. [Configuration](#configuration) is advisory, and `/ork:setup` asks before writing
  268. anything.
  269. ---
  270. ## Install
  271. ```bash
  272. /plugin install ork
  273. ```
  274. No tiering. No version confusion. Just one powerful plugin.
  275. Not on Claude Code? Pull a **starter 12** into any agent (Cursor, Codex, OpenCode, …) via [skills.sh](https://www.skills.sh/yonatangross/orchestkit) — do not install the whole firehose on day one. Picked from measured skills.sh installs on our listing plus the ork skills we actually retain:
  276. | skill | why |
  277. |---|---|
  278. | `devops-deployment` | most-installed cloud skill on our listing |
  279. | `responsive-patterns` | most-installed design skill on our listing |
  280. | `architecture-decision-record` | writing-bucket install leader |
  281. | `ui-components` | design install co-leader |
  282. | `rag-retrieval` | SE install leader |
  283. | `agent-orchestration` | agent-workflow install leader |
  284. | `brainstorm` | our most-retained ork skill |
  285. | `expect` | browser verification we run day to day |
  286. | `auto` | front-door router we actually invoke |
  287. | `review-pr` | ship-loop review we use |
  288. | `fix-issue` | debug loop we use |
  289. | `commit` | ship-loop commit (has a with/without receipt) |
  290. ```bash
  291. npx skills add yonatangross/orchestkit -s devops-deployment -s responsive-patterns -s architecture-decision-record -s ui-components -s rag-retrieval -s agent-orchestration -s brainstorm -s expect -s auto -s review-pr -s fix-issue -s commit
  292. ```
  293. All skills: `npx skills add yonatangross/orchestkit`. Live install totals stay on the [skills.sh badge](https://www.skills.sh/yonatangross/orchestkit) above, not hard-coded here.
  294. ### Cursor
  295. Cursor loads [Agent Plugins](https://agent-plugins.org) and Cursor plugins.
  296. This repo already ships the Agent Plugins manifest at `plugins/ork/plugin.json`.
  297. Add the GitHub repo as a Cursor marketplace (Settings → `yonatangross/orchestkit`),
  298. enable **`ork`**, then **open a new chat**. That is the same plugin Claude Code
  299. installs, not a five-skill fork.
  300. It also ships 14 rules under the plugin's `rules` key, generated from
  301. `src/rules/` and `src/shared/rules/`. They are agent-fetched, so a rule costs
  302. context only when its description matches the task.
  303. Claude hook scripts are not registered for Cursor: they depend on
  304. `${CLAUDE_PLUGIN_ROOT}` (orchestkit#293, closed). Cursor enforcement for HQ
  305. repos stays in the consuming project's `.cursor/hooks.json`.
  306. "Include third-party Plugins" can leak SKILL.md from `~/.claude/plugins`. That
  307. is not an install. Proof is the `ork` plugin id plus the full skill catalog.
  308. ### Codex
  309. Codex uses its own plugin format, skill picker, and standalone role
  310. configuration. Add OrchestKit's Codex marketplace, then install the small
  311. portable workflow pack:
  312. ```bash
  313. codex plugin marketplace add yonatangross/orchestkit --ref main --sparse .agents/plugins --sparse plugins/ork-codex
  314. codex plugin add ork-codex@orchestkit-codex
  315. ```
  316. Restart Codex after installation. Invoke a workflow explicitly with
  317. `$ork-brainstorm`, `$ork-explore`, `$ork-implement`, `$ork-assess`, `$ork-verify`,
  318. or `$ork-review-pr`; their narrow descriptions also let Codex select the relevant
  319. workflow automatically.
  320. The plugin intentionally ships roles as templates because Codex loads custom
  321. roles from `~/.codex/agents/`, not from a plugin manifest. From an OrchestKit
  322. checkout, run this one-time, non-overwriting install:
  323. ```bash
  324. plugins/ork-codex/scripts/install-codex-roles.sh ~/.codex/agents
  325. ```
  326. It installs `ork_explorer`, `ork_implementer`, `ork_reviewer`, and
  327. `ork_verifier`; restart Codex before spawning them.
  328. #### Unattended runs: the `ork-mech` profile
  329. For mechanical work (renames, bumps, codemods, sweeps that end in a diff),
  330. install the shipped profile and run `codex exec` against it:
  331. ```bash
  332. plugins/ork-codex/scripts/install-codex-profile.sh ~/.codex
  333. codex exec --profile ork-mech "<task>" </dev/null
  334. ```
  335. The profile is a FILE, not a snippet you paste into `config.toml`. Measured on
  336. codex-cli 0.153.4: `--profile <name>` layers `$CODEX_HOME/<name>.config.toml`
  337. over the base config, and a legacy `[profiles.<name>]` table left inside
  338. `config.toml` makes the same flag a hard config-load error. The installer
  339. refuses to run next to that table, and refuses to overwrite a profile you
  340. already have.
  341. What it sets, as `codex exec` prints it in its own header:
  342. ```
  343. approval: never
  344. sandbox: workspace-write [workdir, /tmp, $TMPDIR] (network access enabled)
  345. reasoning effort: high
  346. ```
  347. It deliberately does not pin a model (pass `-m`) and does not use
  348. `--dangerously-bypass-approvals-and-sandbox`, which drops the sandbox entirely.
  349. Two contracts a TOML file cannot express, so they stay on the command line:
  350. - **`</dev/null`.** `codex exec` reads stdin even when a prompt argument is
  351. given. An inherited open pipe blocks the run with `Reading additional input
  352. from stdin...` and no timeout.
  353. - **`--add-dir` inside a git worktree.** The writable roots are
  354. `[workdir, /tmp, $TMPDIR]`. A linked worktree's git common dir sits outside
  355. the workdir, so the first commit dies on `index.lock`. Add it:
  356. ```bash
  357. codex exec --profile ork-mech \
  358. --add-dir "$(git rev-parse --path-format=absolute --git-common-dir)" \
  359. "<task>" </dev/null
  360. ```
  361. #### Keeping the install current
  362. `ref main` in the marketplace source is a cached snapshot, not a tracker. On the
  363. 2026-09-08 audit machine `codex plugin list` showed `10.0.0-beta.5` while main
  364. was three releases ahead. After an OrchestKit release, update and check:
  365. ```bash
  366. codex plugin update
  367. codex plugin list | grep ork-codex # installed version
  368. jq -r .version plugins/ork-codex/.codex-plugin/plugin.json # what main ships
  369. ```
  370. #### Documentation lookup (context7)
  371. The plugin ships a [context7](https://context7.com) MCP server in its own
  372. manifest (`mcpServers` in `.codex-plugin/plugin.json`, defined in `mcp.json`),
  373. so installing the plugin registers it. Confirm with `codex mcp get context7`.
  374. It is scoped to the only two tools context7 exposes, `resolve-library-id` and
  375. `query-docs`, and it uses the hosted HTTP transport rather than an `npx` stdio
  376. child, so it costs no extra process per Codex session.
  377. Export a key before starting Codex. The plugin references the variable name
  378. and never stores the value, so no token is written to `~/.codex/config.toml`:
  379. ```bash
  380. export CONTEXT7_API_KEY_CODEX="<your-context7-api-key>"
  381. ```
  382. Put that in your shell profile so every Codex session inherits it. Get the
  383. key from your own context7 account and keep the value out of the repository.
  384. If you store it in a secret manager, substitute your own vault and item names
  385. (with the 1Password CLI the reference is `op://<vault>/<item>/credential`), and
  386. cache the resolved value instead of re-reading the vault in every shell: each
  387. raw read is a separate unlock prompt.
  388. Two behaviors worth knowing:
  389. - A server you already define yourself under `[mcp_servers.context7]` in
  390. `~/.codex/config.toml` **wins**, and the plugin's definition is ignored
  391. entirely (including its tool scoping). That is intentional: your own
  392. configuration is never overridden. Remove your entry if you want the
  393. plugin's.
  394. - Without a valid key the server still connects and still lists its tools.
  395. Only a real call fails, with `Invalid API key`. A successful connection is
  396. therefore not proof of authentication.
  397. ### pi
  398. pi (0.85) reads the same SKILL.md format, and the repo now carries a `pi`
  399. manifest, so the package installs directly:
  400. ```bash
  401. pi install git:github.com/yonatangross/orchestkit
  402. ```
  403. That registers every skill. Add `-l` to write `.pi/settings.json` in the
  404. project instead of your user settings. Pointing pi at a checkout still works
  405. and needs no install (`pi --skill ./plugins/ork/skills`).
  406. The 29 skills marked `disable-model-invocation` stay reachable only as
  407. `/skill:<name>`. Two measured caveats: `--no-builtin-tools` hides every skill
  408. (pi lists skills only when a file-reading tool is enabled), and `pi -p` blocks
  409. on an open stdin, so headless runs need `</dev/null`.
  410. MCP servers for pi come from `.pi/mcp.json`, then `.mcp.json`, then
  411. `~/.config/mcp/mcp.json`. Copy the shipped template to get the recommended
  412. servers with a read-only `includeTools` allowlist per server:
  413. ```bash
  414. cp .pi/mcp.json.example .pi/mcp.json
  415. ```
  416. Full detail and the tracking epic:
  417. [OrchestKit on pi, Codex and Cursor](https://orchestkit.dev/docs/guides/orchestkit-on-pi-codex-cursor).
  418. ### Devin
  419. ```bash
  420. devin plugins install yonatangross/orchestkit
  421. ```
  422. Devin checks a plugin root for a manifest in this order: `.devin-plugin/plugin.json`,
  423. then `.claude-plugin/plugin.json`, then root `plugin.json`
  424. ([plugins reference](https://docs.devin.ai/cli/extensibility/plugins/overview)).
  425. The repo root only had the last one, the [Agent Plugins](https://agent-plugins.org)
  426. manifest, which fixes skills at a root `skills/` directory this repo does not have,
  427. so a bare install used to list 0 of 107 skills, 0 hooks, and only the `AGENTS.md`
  428. rule (GH-4146). A root level `.devin-plugin/plugin.json` now maps `skills` to
  429. `./plugins/ork/skills`, the same built tree Claude Code and Cursor already read.
  430. `devin plugins info ork` lists 76 of 107 skills after that change (measured on
  431. Devin CLI 3000.10.27). The other 31, `auto`, `verify`, `help`, `brainstorm`, and
  432. similar router or workflow skills, declare a `triggers:` frontmatter key shaped
  433. as an object (`keywords`, `examples`, `anti-triggers`). Devin's own `triggers`
  434. field expects a flat `[user, model]` list, and the shape mismatch drops the
  435. whole skill instead of warning. Renaming that key touches three other readers
  436. (`scripts/eval/eval-coverage.sh`, `tests/skills/triggering/test-trigger-keywords.sh`,
  437. and the `SKILL_ONLY` allowlist in `tests/plugins/test-command-frontmatter-passthrough.sh`),
  438. so it stays tracked on GH-4146 instead of folded into this fix. Installing the
  439. subpath directly, `devin plugins install yonatangross/orchestkit#plugins/ork`,
  440. reaches the same 76 skills plus the plugin's 36 custom subagents, which neither
  441. path surfaces through `devin plugins info`. That subpath is what this repo's own
  442. `.claude-plugin/marketplace.json` already points Claude Code at.
  443. OrchestKit's plugin hooks work under Claude Code only today; none of the <!--ork:hooks-->163<!--/ork-->
  444. fire under Devin. Two reasons. Location: OrchestKit ships hooks at
  445. `plugins/ork/hooks/hooks.json`, while Devin's own plugin format reads a bare
  446. `hooks.json` at the plugin root instead. Shape and tool names: OrchestKit's
  447. file matches Claude Code tool names such as `Bash` and `Write|Edit` and
  448. expands `${CLAUDE_PLUGIN_ROOT}` in command args, both Claude Code specific.
  449. Devin's own
  450. [hooks.v1.json](https://docs.devin.ai/cli/extensibility/hooks/overview) format
  451. matches its own tool names instead, `exec`, `write`, `edit` and friends (see
  452. [lifecycle hooks](https://docs.devin.ai/cli/extensibility/hooks/lifecycle-hooks)),
  453. inside the same `{matcher, hooks:[{type, command}]}` event map shape.
  454. `devin plugins info` confirms `Hooks (none)` even once skills resolve.
  455. `PreToolUse` and `PostToolUse` read the closest to what OrchestKit's own
  456. pretool and posttool hooks already do, and are the first candidates for a
  457. future Devin hook manifest once the tool name mapping is written deliberately
  458. instead of guessed.
  459. Pin a release instead of tracking `main`: release-please tags every release as
  460. `v10.0.0-beta.N`. The single argument `devin plugins install <source>` command
  461. has no ref pinning syntax of its own; pin through a `requiredPlugins` entry
  462. (in this repo's own `.devin/config.json`, or a personal, org, or enterprise
  463. manifest) instead:
  464. ```json
  465. {
  466. "requiredPlugins": [
  467. { "source": "github", "repo": "yonatangross/orchestkit", "ref": "v10.0.0-beta.30" }
  468. ]
  469. }
  470. ```
  471. ### Antigravity
  472. ```bash
  473. npx skills add yonatangross/orchestkit -s devops-deployment -s responsive-patterns -s architecture-decision-record -s ui-components -s rag-retrieval -s agent-orchestration -s brainstorm -s expect -s auto -s review-pr -s fix-issue -s commit
  474. ```
  475. Antigravity (`agy`, Google's agentic CLI) reads Agent Skills from a workspace
  476. `.agents/skills` directory, so the shared skills.sh line above is the install
  477. path and Antigravity is already in that installer's universal target list
  478. (measured on agy 1.2.6; full evidence in
  479. `docs/audits/agy-host-support-2026-09-18.md`). Verify from the repo root with
  480. `agy -p "/skills" --add-dir "$PWD"`: print mode only registers a workspace from
  481. an absolute `--add-dir`, while the interactive TUI takes the launch directory
  482. as the workspace and needs no flag.
  483. Two honest gaps. `npx skills add -g` lands in `~/.agents/skills`, which agy
  484. does not read; user scope lives at `~/.gemini/config/skills` (or
  485. `~/.gemini/skills`). And none of ork's <!--ork:hooks-->163<!--/ork--> hooks port: agy's `hooks.json` uses
  486. its own event names and payload shape, and `agy plugin validate plugins/ork`
  487. reports hooks skipped because ork keeps them at `hooks/hooks.json`, not the
  488. plugin-root `hooks.json` agy looks for.
  489. A fuller install exists but needs a local checkout:
  490. `agy plugin validate <checkout>/plugins/ork` reports `skills: 108 processed`,
  491. `agents: 36 processed`, and `agy plugin install <checkout>/plugins/ork` copies
  492. them into `~/.gemini/config/plugins/ork/` namespaced `ork:<name>`, which matches
  493. the `/ork:<skill>` spelling. Plugin-bundled `mcp_config.json` also works
  494. (marker-verified end to end), but a repo-level `.mcp.json` is never read.
  495. ---
  496. ## FAQ
  497. <details>
  498. <summary><strong>Plugin not found?</strong></summary>
  499. ```bash
  500. /plugin list
  501. /plugin uninstall ork && /plugin install ork
  502. ```
  503. </details>
  504. <details>
  505. <summary><strong>Hooks not firing?</strong></summary>
  506. Run `/ork:doctor` to diagnose.
  507. </details>
  508. <details>
  509. <summary><strong>Claude Code version?</strong></summary>
  510. Requires **≥2.1.277** (supported floor; Opus 5 as the default Opus, Opus 5.5 from 2.1.280, `xhigh` effort, dynamic workflows, `sandbox.network.strictAllowlist`, native binary, hardened `Bash(rm:*)`/`Bash(find:*)` rules). Check with `claude --version`.
  511. Raising this floor is a breaking change and ships as a major release. See [STABILITY.md](STABILITY.md) for the full contract, and `shared/cc-support.json` for the authoritative window.
  512. </details>
  513. <details>
  514. <summary><strong>Superpowers vs OrchestKit?</strong></summary>
  515. Complementary, not a rival listing. Superpowers (official Anthropic marketplace) is process: how the agent works a task. OrchestKit is production patterns plus lifecycle hooks. Honest split: [docs](https://orchestkit.dev/docs/getting-started/superpowers) · [yonyon.ai](https://yonyon.ai/compare/orchestkit-superpowers).
  516. </details>
  517. ---
  518. ## Development
  519. ```bash
  520. npm run build # Build plugins from src/
  521. npm test # Run all tests
  522. ```
  523. Edit `src/` and `manifests/`, never `plugins/` (generated).
  524. See [CONTRIBUTING.md](./CONTRIBUTING.md) for details.
  525. ---
  526. ## What's New
  527. <!--ork:whats-new-->
  528. <!-- AUTO-GENERATED from CHANGELOG.md by scripts/stamp-whats-new.mjs — do not hand-edit between the ork:whats-new markers. -->
  529. <!-- Regenerated on `npm run build`; CI (`--check`) fails if this is stale. Full history: [CHANGELOG.md](CHANGELOG.md). -->
  530. **[v10.14.0](https://github.com/yonatangross/orchestkit/compare/v10.13.0...v10.14.0)** · 2026-11-17
  531. - **cli:** orchestkit install finds the coding agents on this machine (#5267)
  532. - **codex:** add ork-commit and ork-create-pr to the Codex pack (#5261)
  533. - **fix-issue:** safety-fact proof ladder and review-pr proof criterion (#5273)
  534. - **hooks:** record a fixed error_class on PostToolUseFailure telemetry (#5248)
  535. - **review-pr:** proven-by-a-run tier in review-fanout (#5274)
  536. - …and 27 more (see [CHANGELOG.md](CHANGELOG.md))
  537. **[v10.13.0](https://github.com/yonatangross/orchestkit/compare/v10.12.0...v10.13.0)** · 2026-11-15
  538. - **ci:** scheduled whole-tree npm audit with a cannot-audit item (#5230)
  539. - **plugins:** move 26 reference skills to the opt-in ork-extras plugin (#5220)
  540. - **ci:** make the nightly health report prove its Claude call (#5221)
  541. - **ci:** the npm audit gate judges the change, not the tree (#5233)
  542. - **disk-cleanup:** ask before images, keep new snapshots, pin apply (#5213)
  543. - …and 11 more (see [CHANGELOG.md](CHANGELOG.md))
  544. **[v10.12.0](https://github.com/yonatangross/orchestkit/compare/v10.11.0...v10.12.0)** · 2026-11-12
  545. - **hooks:** ship 2 hook processes per tool call, not 3 (#5214)
  546. - **skills:** add /ork:disk-cleanup (#5211)
  547. - **cursor:** commands run skill scripts only from the installed plugin (#5216)
  548. - **docs:** hook pages say when the shipped plugin skips an event (#5215)
  549. - **docs:** try Chromium before apt, so the mirror is not on every build (#5212)
  550. - …and 5 more (see [CHANGELOG.md](CHANGELOG.md))
  551. **[v10.11.0](https://github.com/yonatangross/orchestkit/compare/v10.10.8...v10.11.0)** · 2026-11-10
  552. - **cc:** adopt Claude Code 2.1.310 through 2.1.322 (#5181)
  553. - **demo-producer:** add code-drawn mode, storyboard renders the film (#5185)
  554. - **docs-site:** count page views cookieless before consent (#5180)
  555. - **docs-site:** make the consent banner easier to notice (#5179)
  556. - **agents:** raise builder maxTurns above documented workload (#5192)
  557. - …and 11 more (see [CHANGELOG.md](CHANGELOG.md))
  558. **[v10.10.8](https://github.com/yonatangross/orchestkit/compare/v10.10.7...v10.10.8)** · 2026-11-09
  559. - **release:** fetch mcp-publisher through the authenticated API (#5162)
  560. **[v10.10.7](https://github.com/yonatangross/orchestkit/compare/v10.10.6...v10.10.7)** · 2026-11-09
  561. - **ci-sentinel:** match the skill to the locked-down dispatch (#5156)
  562. - **ci-sentinel:** say every failed check, and pin the playbook copy (#5159)
  563. - **deps-dev:** bump @biomejs/biome (#5153)
  564. - **deps-dev:** bump @sanity/client (#5154)
  565. - **deps:** bump @anthropic-ai/sdk (#5151)
  566. - …and 4 more (see [CHANGELOG.md](CHANGELOG.md))
  567. **[v10.10.6](https://github.com/yonatangross/orchestkit/compare/v10.10.5...v10.10.6)** · 2026-11-09
  568. - **ci-sentinel:** refresh drifted skill text (#5147)
  569. - **ci:** remove the ci-sentinel ceiling that could never fire (#5141), closes [#5138](https://github.com/yonatangross/orchestkit/issues/5138)
  570. - **ci:** tell claude-triage the repo so it stops running out of turns (#5143), closes [#5142](https://github.com/yonatangross/orchestkit/issues/5142)
  571. - **docs:** serve the server card media type when a client asks (#5125)
  572. - **secrets-veil:** keep image URLs on named media hosts readable (#5146)
  573. - …and 4 more (see [CHANGELOG.md](CHANGELOG.md))
  574. **[v10.10.5](https://github.com/yonatangross/orchestkit/compare/v10.10.4...v10.10.5)** · 2026-11-08
  575. - Codex hook paths and docs MCP query validation (#5119)
  576. - **hooks:** state machine audit bundle (#5117)
  577. - **skills:** bill long Haiku 5.5 prompts on the higher rate card (#5124)
  578. - **skills:** send the event name from fork skill body loaders (#5122)
  579. - **skills:** skill quality round 5, pass the fitness checklist (#5104)
  580. _See [CHANGELOG.md](CHANGELOG.md) for the full release history._
  581. <!--/ork-->
  582. ---
  583. ## Community
  584. Join the **Building with AI** community for AI dev tips, OrchestKit support, and connecting with other builders:
  585. | Room | Who it's for | Link |
  586. |------|--------------|------|
  587. | **Building with AI** | The umbrella community. One join, every room below. | [Join](https://platform.yonyon.ai/circle?ref=readme) |
  588. | **Builders** | For people already building | [Join](https://yonyon.ai/go/builders?utm_campaign=readme) |
  589. | **OrchestKit** | For OrchestKit users | [Join](https://yonyon.ai/go/orchestkit?utm_campaign=readme) |
  590. | **AI for Business** | For people leading AI adoption | [Join](https://yonyon.ai/go/business?utm_campaign=readme) |
  591. Names and audiences match what [yonyon.ai](https://yonyon.ai/en) renders, so the two surfaces cannot drift. Every link resolves through `yonyon.ai/go/*`, so a rotated invite never needs a README change and no raw invite is published here.
  592. ---
  593. ## OrchestKit family
  594. - **[Tower](https://github.com/yonatangross/tower)**: a macOS menu bar app for people who run several AI coding subscriptions: usage per account, room left, when to switch. Also an MCP server (`tower mcp`).
  595. ---
  596. ## Who builds this
  597. OrchestKit is built and maintained by **[Yonatan Gross](https://github.com/yonatangross)** — [Yonyon AI](https://yonyon.ai/en), an AI consulting practice. It is the toolkit extracted from real client work, not a side project: the patterns here are the ones that survived shipping.
  598. It stays MIT and free. Nothing is gated, and none of the below changes that.
  599. **Working out where AI actually fits in your business?** The [**AI readiness audit**](https://platform.yonyon.ai/ai-audit) is a free assessment that maps your workflows and returns a prioritized report — the same diagnostic that opens a consulting engagement.
  600. **Want the toolkit running properly in your team?** Setup, configuration, and a working agent loop tailored to your stack is something I do as a fixed-scope engagement. Start a [discussion](https://github.com/yonatangross/orchestkit/discussions) or reach out through the [community](https://yonyon.ai/go/business?utm_campaign=readme).
  601. Security policy and reporting: [SECURITY.md](SECURITY.md).
  602. ---
  603. <div align="center">
  604. **[Docs](https://orchestkit.dev/)** · **[Issues](https://github.com/yonatangross/orchestkit/issues)** · **[Discussions](https://github.com/yonatangross/orchestkit/discussions)** · **[Community](https://platform.yonyon.ai/circle?ref=readme)**
  605. MIT License · [@yonatangross](https://github.com/yonatangross)
  606. </div>

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.