{
"markdown": "# Artibot\n\n[](plugins/artibot/CHANGELOG.md)\n[](LICENSE)\n[](package.json)\n[](plugins/artibot/tests/)\n[](plugins/artibot/tests/)\n[](plugins/artibot/eslint.config.js)\n[](https://github.com/anthropics/claude-code)\n[](https://claude.com/cowork)\n[](plugins/artibot/lib/runtime/middleware/subagents.js)\n\n> **Cognitive orchestration OS for Claude Code with hierarchical memory, RLVR self-learning, MCP server, and multi-platform agent teams.**\n\nThis repository ships **two complementary plugins** under one marketplace:\n\n| Plugin | Target | Version | Best for |\n|---|---|---|---|\n| [`artibot`](./plugins/artibot/) | Claude Code (developer CLI) | **4.56.0** | full Agent Teams orchestration, TDD, code review, security audits, RLVR learning, MCP server, **Goal-driven autopilot**, **`/learning` diagnostics**, **`/save` + `/resume` single-shot handoff**, **`/go` → `/orchestrate` build-sequence hand-off**, **ambient conversation ledger (no-command capture)**, **safety boost (machineId frontmatter, git-lock fail, 10m throttle)** |\n| [`artibot-cowork`](./plugins/artibot-cowork/) | Claude Cowork (knowledge workers) | **3.1.0** | marketing campaigns, long-form writing, AEO/GEO content, KR-market SEO, AI-slop detection, **Claude Design, Routines, Ultraplan, Monitor** |\n\nBoth plugins share the same DEV protocol, Korean market expertise, data-sovereignty policy, and 6-stage content quality pipeline. They differ only in **target environment** and **skill mix**.\n\nA cognitive orchestration plugin for [Claude Code](https://github.com/anthropics/claude-code) powered by **dual-process routing**, **lifelong learning**, and the **native Agent Teams API**. Artibot uses a System 1/2 cognitive architecture to classify every request, assembles specialized agent teams with P2P communication, and continuously learns from session outcomes to improve routing accuracy over time.\n\n
\n \n
\n\n## Overview\n\nMost Claude Code plugins use simple sub-agent (unnamed, fire-and-forget) delegation -- fire-and-forget with one-way result reporting. Artibot takes a fundamentally different approach by using Claude's **native Agent Teams API** as its core engine, enabling true team orchestration:\n\n| Capability | Sub-Agent (Task) | Agent Teams (Artibot) |\n|------------|------------------|----------------------|\n| Communication | Return result to parent only | P2P bidirectional messaging (SendMessage) |\n| Task Management | Parent manages everything | Shared task list (TaskCreate/TaskList) |\n| Self-Assignment | Not possible | Teammates self-claim from task list |\n| Peer Communication | Not possible | Direct DM + broadcast |\n| Plan Approval | Not possible | plan_approval_response |\n| Lifecycle | One-shot | Create -> Work -> Shutdown -> Cleanup |\n\n## Key Features\n\n- **Cognitive Architecture** -- Dual-process System 1/2 routing inspired by Kahneman's theory: fast intuitive responses for simple tasks, deep deliberative reasoning for complex ones\n- **Lifelong Learning** -- GRPO-based batch learning from session outcomes with automatic knowledge transfer between System 1 and System 2 caches\n- **Decision Ledger (v5 Observe)** -- Append-only records of routing, mission, human-gate and verification decisions, written as the decision happens. Recording only: runtime behavior is unchanged and routing receipts carry `applied:false`. Read them back with `/scorecard --session|--routing` and `/doctor` Check 8/9\n- **CTO-Led Orchestration** -- `orchestrator` agent leads 30 specialist agents as a team coordinator (delegation mode: no direct coding)\n- **Intelligent Delegation** -- Auto-selects Sub-Agent (simple) vs Agent Team (complex) based on cognitive complexity scoring\n- **5 Orchestration Patterns** -- Leader, Council, Swarm, Pipeline, Watchdog\n- **8 DAG Playbooks** -- Feature, Bugfix, Refactor, Security + Marketing Campaign, Marketing Audit, Content Launch, Competitive Analysis (DAG-based with parallel node execution, topological sort, and cycle detection)\n- **79 Slash Commands** -- `/sc` smart router, `/save`, `/resume`, `/daily`, `/team`, `/orchestrate`, `/spawn`, `/implement`, `/visual-check`, `/sc playbook`, `/learning`, `/ultrareview`, `/ultraplan`, `/audit-claude-md`, `/export`, and more\n- **30 Specialized Agents** -- Architecture, security, frontend, backend, testing, DevOps, marketing, SEO, analytics, investigation, audit, and more (20 on the `opus` tier, 10 design/review/judge agents on the `fable` tier)\n- **114 Domain Skills** -- 11 persona skills, 8 core skills, 16 language skills, 8 utility skills, 35 marketing skills, visual-validation, daily, team, session-worklog, vibe-coding, repo-benchmarking, git-worktree, dynamic-context-injection, claude-md-auditor, skill-authoring (all enhanced with Anthropic best-practice descriptions, workflow checklists, HITL v2 conversational checkpoints, output templates, and freedom levels)\n- **10 Auto-Activating Rules** -- DEV protocol, quality gates, agent coordination, config safety, frontend/backend/test patterns, clean state enforcement, verification discipline, question recommendations\n- **Guard Registry** -- Centralized guard pipeline with `registerGuard()`/`executeChain()` API, 6 built-in guards extracted from hook scripts (75% code reduction)\n- **Loop Detection** -- Circular buffer-based agent loop detection with fingerprint matching, automatic warn/block on repeated tool calls\n- **Clean State Enforcement** -- TaskCompleted hook ensures lint+test verification at feature completion boundaries\n- **27 Hook Registrations** -- Across 16 event types: cognitive routing, lifelong learning, session lifecycle, dangerous command blocking, auto-formatting, team tracking, loop detection, clean state checks, HTTP webhook notifications, git autopilot\n- **Advisory File Locking** -- Spin-lock based file locking for concurrent hook state access, fail-open pattern prevents workflow blocking\n- **DEV Protocol** -- Mandatory Decompose-Execute-Verify workflow with zero-skip policy for all code changes\n- **Vibe Coding Support** -- Natural language request handling with read-first, verify-after, evidence-based completion\n- **Visual Validation Pipeline** -- SSIM-based screenshot comparison, auto-fix suggestion, iterative correction loop via Playwright MCP\n- **Conversation-to-Memory** -- Auto-extracts rules and decisions from user messages (Korean/English), injects into skills dynamically\n- **Project CLAUDE.md Seeding** -- `install.sh` auto-generates project-level CLAUDE.md with Artibot methodology and DEV protocol\n- **Runtime Middleware Pipeline** -- 11-stage middleware engine (lifecycle, router, memory, skills, tasks, subagents, guardrail, summarization, token-usage, checkpoint, cache-roi) for runtime context injection\n- **Output Design System** -- 7 output styles with design tokens (tokens.md) and narrative template for consistent formatting\n- **Forked Context Skills** -- All 114 skills run in isolated forked context for clean execution without cross-contamination\n- **HTTP Webhook Notifications** -- Session events sent to Slack, Discord, or custom endpoints via configurable webhooks\n- **Cross-Platform Compatible** -- Works with Gemini CLI, OpenAI Codex, and Cursor via platform adapters\n- **Zero Dependencies** -- Pure Node.js built-in modules only (`node:fs`, `node:path`, `node:os`)\n\n## Get Started\n\n### Prerequisites\n\n**Agent Teams** is auto-enabled by Artibot on first session start. No manual setup needed.\n\n### Installation\n\n**Recommended: native marketplace (no clone)** — run inside any Claude Code session:\n```text\n/plugin marketplace add Yoodaddy0311/artibot\n/plugin install artibot@artibot\n```\n\nClaude Code fetches the plugin straight from GitHub into its plugin cache — no\n`git clone`, no shell script. Commands, agents, skills, and hooks load via\n`${CLAUDE_PLUGIN_ROOT}`; updates come through `/plugin marketplace update artibot`.\nNative install **namespaces every command** under the `artibot:` prefix (call\n`/artibot:save`, `/artibot:sc`). Agent Teams auto-enables on first session start.\nTo uninstall: `/plugin uninstall artibot@artibot`.\n\n> **Native install does not deliver the 10 auto-activating rules** (DEV Protocol,\n> Quality Gates, agent-coordination, config-safety, clean-state, frontend/backend/test\n> patterns, verification-discipline, question-recommendations). Claude Code's plugin schema has no `rules` field, so they load only\n> when `install.sh` copies them to `~/.claude/rules/artibot/`. If you want that\n> automatic DEV-protocol / quality-gate enforcement, use the full install below.\n> `/theme` and the themed statusline now auto-resolve both layouts (native install\n> included); a native-only install may need `/theme` re-run after a major `/plugin`\n> update. `/update` still assumes the flat layout and is being migrated. Commands,\n> agents, skills, and hooks themselves work identically on both paths.\n\n**Full install: `install.sh` (flat commands, no prefix)**\n```bash\ngit clone https://github.com/Yoodaddy0311/artibot.git\ncd artibot/plugins/artibot\nbash install.sh # macOS / Linux / Git Bash on Windows\n```\n\nThis flat-copies agents and commands into `~/.claude/`, so slash commands are\ncalled **without a prefix** (`/save`, `/sc`, `/daily`). It enables Agent Teams,\nwires the themed statusline, and seeds a conservative read-only permission\nallowlist (`Read`/`Glob`/`Grep`) into `~/.claude/settings.json`, removing the\nrepeated approval prompts new users hit on first run. To uninstall:\n`bash install.sh uninstall`.\n\n> **Windows:** run `install.sh` from **Git Bash** ([Git for Windows](https://gitforwindows.org/)),\n> or use the native PowerShell installer: `powershell -ExecutionPolicy Bypass -File install.ps1`.\n\n**Command invocation differs by install method:**\n\n| Install method | Command form | Example |\n|---|---|---|\n| Native marketplace (`/plugin install`) | namespaced `artibot:` | `/artibot:save`, `/artibot:sc` |\n| `install.sh` / `install.ps1` (flat) | flat (no prefix) | `/save`, `/sc`, `/daily` |\n\n### Requirements\n\n- [Claude Code](https://github.com/anthropics/claude-code) CLI\n- Node.js >= 18.0.0\n- Agent Teams (auto-enabled by Artibot, or manually set `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`)\n\n### Cross-Platform Installation\n\nArtibot works beyond Claude Code. Built-in adapters auto-convert agents, skills, and commands for each platform.\n\n> Replace `` with the root directory path of the target project.\n\n| Platform | Compatibility | Agent Teams | Sub-Agent | Skills |\n|----------|:------------:|:-----------:|:---------:|:------:|\n| Claude Code | 10/10 | ✅ | ✅ | ✅ |\n| Gemini CLI | 9/10 | ❌ | ✅ | ✅ |\n| Codex CLI | 8/10 | ❌ | ✅ | ✅ |\n| Antigravity | 8/10 | ❌ | ✅ | ✅ |\n| Cursor IDE | 6/10 | ❌ | ⚠️ | ✅ |\n\n#### Gemini CLI\n\n```bash\ngit clone https://github.com/Yoodaddy0311/artibot.git\ncd artibot\nnode --input-type=module -e \"import { exportForGemini } from './plugins/artibot/lib/adapters/skill-export.js'; const r = await exportForGemini({ pluginRoot: './plugins/artibot' }); console.log(r.summary);\"\n```\n\n#### OpenAI Codex CLI\n\n```bash\ngit clone https://github.com/Yoodaddy0311/artibot.git\ncd artibot\nnode --input-type=module -e \"import { exportForCodex } from './plugins/artibot/lib/adapters/skill-export.js'; const r = await exportForCodex({ pluginRoot: './plugins/artibot' }); console.log(r.summary);\"\n```\n\n#### Google Antigravity\n\n```bash\ngit clone https://github.com/Yoodaddy0311/artibot.git\ncd artibot\nnode --input-type=module -e \"import { exportForGemini } from './plugins/artibot/lib/adapters/skill-export.js'; const r = await exportForGemini({ pluginRoot: './plugins/artibot' }); console.log(r.summary);\"\n```\n\n> **Note:** Agent Teams API is exclusive to Claude Code. Other platforms automatically fall back to Sub-Agent mode (one-way delegation). Learning, memory, and swarm intelligence work identically across all platforms.\n>\n> For the full cross-platform guide including Cursor IDE and batch export, see [plugins/artibot/README.md](plugins/artibot/README.md).\n\n## What can Artibot do?\n\n### `/implement` -- Feature Implementation Pipeline\n\n**Before (manual):**\n```\nYou: \"implement user authentication\"\nClaude: writes code in one file, misses edge cases, no tests, no review\nYou: manually request tests, then review, then fix issues found in review\n-> 4+ back-and-forth prompts, inconsistent coverage\n```\n\n**After (with Artibot):**\n```\nYou: \"implement user authentication\"\nArtibot: /implement triggers automatically\n -> planner agent: decomposes into 5 subtasks\n -> architect agent: designs auth flow + data model\n -> backend-developer + frontend-developer: implement in parallel (Agent Teams)\n -> tdd-guide agent: writes tests (80%+ coverage)\n -> code-reviewer agent: reviews with 4-severity classification\n -> result: complete feature with tests, reviewed, ready to merge\n-> 1 prompt, full pipeline\n```\n\n### `/team` -- Parallel Team Execution\n\n**Before (manual):**\n```\nYou: \"refactor the payment module and update the API docs\"\nClaude: does refactoring first (10 min), then docs (5 min) = 15 min sequential\nNo cross-check between the two tasks\n```\n\n**After (with Artibot):**\n```\nYou: \"refactor the payment module and update the API docs\"\nArtibot: /team triggers automatically\n -> refactor-cleaner agent: refactoring (teammate A)\n -> doc-updater agent: documentation (teammate B)\n -> both run in parallel via Agent Teams API (P2P messaging)\n -> cross-check phase: each reviews the other's output\n -> result: 7 min total, cross-verified\n```\n\n### `/export` -- Cross-Platform Export\n\n**Before (manual):**\n```\nYou want to use Artibot agents in Cursor IDE\n-> manually rewrite each agent .md file to .mdc format\n-> figure out Cursor Rules-for-AI conventions\n-> repeat for 30 agents = hours of tedious work\n```\n\n**After (with Artibot):**\n```\n/export cursor\n -> auto-converts all 30 agents to .mdc format\n -> places them in .cursor/rules/ with correct frontmatter\n -> preserves agent expertise and tool permissions\n -> supports: cursor, codex, opencode, antigravity, all\n-> 1 command, all platforms\n```\n\n---\n\n## Usage\n\n### Smart Routing\n\nThe `/sc` command analyzes natural language intent and routes to the optimal command:\n\n```\n/sc implement login feature\n-> routes to /implement -> spawns named planner + architect + developer + reviewer\n```\n\n```\n/sc analyze security vulnerabilities in auth module\n-> routes to /analyze --focus security -> delegates to security-reviewer sub-agent\n```\n\n```\n/sc launch email marketing campaign for product launch\n-> routes to /campaign -> spawns named marketing-strategist + data-analyst + ad-specialist\n```\n\n### Direct Commands\n\n```bash\n/implement user authentication API --type api --tdd\n/code-review @src/auth/\n/test --coverage\n/git commit\n/campaign product launch --channels email,social\n/seo audit @website\n```\n\n### Team Orchestration\n\nFor complex tasks, Artibot assembles a full Agent Team:\n\n```bash\n/orchestrate payment system --pattern feature\n```\n\nThis triggers the full team lifecycle:\n1. Fix the run slug `payment-feature` -- no team is created; the session has one implicit team\n2. `Agent(planner, name=\"payment-feature-planner\")` + `Agent(architect, name=\"payment-feature-architect\")` + ... -- spawn teammates\n3. `TaskCreate` per phase (plan -> design -> implement -> review) -- populate task list\n4. `TaskUpdate` -- set dependencies and assign teammates\n5. `SendMessage` -- P2P coordination between teammates\n6. `shutdown_request` to every teammate -- that is the whole cleanup\n\n### Marketing Orchestration\n\n```bash\n/campaign product launch --channels email,social,ads\n```\n\nUses the marketing-campaign playbook:\n1. `[Leader]` marketing-strategist defines strategy\n2. `[Council]` team plans channels and content\n3. `[Swarm]` specialists create content in parallel\n4. `[Council]` review and optimize\n5. `[Leader]` launch coordination\n\n### Parallel Execution\n\n```bash\n/spawn full codebase security audit --mode parallel --agents 5\n```\n\nSpawns 5 teammates that self-claim tasks from the shared task list and report findings via `SendMessage`.\n\n## Architecture\n\n```\n+-------------------------------------------------+\n| User Request |\n+------------------+------------------------------+\n |\n v\n+-------------------------------------------------+\n| /sc Router (Intent Analysis) |\n| keyword 40% + context 40% + flags 20% |\n+------------------+------------------------------+\n |\n v\n+-------------------------------------------------+\n| Delegation Mode (Complexity Scoring) |\n| score < 0.4 -> Sub-Agent | >= 0.4 -> Team |\n+-------+---------------------------------+-------+\n | |\n v v\n+---------------+ +-----------------------+\n| Sub-Agent | | Agent Teams Engine |\n| unnamed | | |\n| one-way | | run slug fixed |\n| | | -> Agent(name=...) |\n| | | -> TaskCreate |\n| | | -> SendMessage(P2P)|\n| | | -> shutdown_request|\n+---------------+ +-----------+-----------+\n |\n v\n +-----------------------+\n | orchestrator (CTO) |\n | Leader | Council | |\n | Swarm | Pipeline | |\n | Watchdog |\n +-----------+-----------+\n |\n v\n +-----------------------+\n | 30 Specialist Agents |\n | TaskList -> self-claim|\n | SendMessage -> P2P |\n | TaskUpdate -> report |\n +-----------------------+\n```\n\n**Decision ledger (v5 Observe).** Alongside this path, an append-only ledger records the\ndecisions it makes -- which model a spawn resolved to, where a hook stopped and asked a human,\nwhat was and was not verified. It observes only: nothing in the flow above is routed, blocked,\nor ordered differently because the ledger exists.\n\n## Cognitive Architecture\n\nArtibot uses a dual-process cognitive model inspired by Daniel Kahneman's System 1/System 2 theory to classify and route every user request:\n\n```\nUser Request\n |\n v\n+-------------+\n| Cognitive | Complexity Score = weighted sum of:\n| Router | steps (0.25) + domains (0.20)\n| (hook-based) | + uncertainty (0.20) + risk (0.20) + novelty (0.15)\n+------+------+\n |\n +--- score < 0.4 ---> System 1 (Fast / Intuitive)\n | - Pattern-matched from cached experience\n | - Target latency: < 100ms\n | - Keyword-based heuristic scoring\n |\n +--- score >= 0.4 ---> System 2 (Deep / Deliberative)\n | - Multi-step structured reasoning\n | - Full context + dependency analysis\n | - Sandbox verification for high-risk ops\n |\n +--- confidence < 0.6 or latency exceeded ---> Escalation\n System 1 -> System 2 automatic fallback\n```\n\n### Escalation Rules\n- System 1 confidence drops below 0.6\n- Processing time exceeds 100ms\n- No matching pattern in System 1 cache\n- Security or production keywords detected\n- Request spans 3+ domains\n- Explicit `--think`, `--think-hard`, or `--ultrathink` flag\n\n### Integration with Delegation\n\nThe cognitive router feeds directly into the orchestration delegation mode:\n\n| Complexity Score | System | Delegation Mode |\n|-----------------|--------|-----------------|\n| < 0.4 | System 1 | Sub-Agent (Task tool) |\n| >= 0.4 | System 2 | Agent Team (Teams API) |\n\n#### Sub-Agent vs Agent Teams\n\n| Aspect | Sub-Agent | Agent Teams |\n|--------|-----------|-------------|\n| **Delegation** | unnamed single call | named `Agent(name=...)` spawns |\n| **UI Visibility** | Hidden (background) | Teammates shown below prompt |\n| **Communication** | One-way (result only) | P2P bidirectional (SendMessage) |\n| **Task Management** | None | Shared task list (TaskCreate/Update/List) |\n| **Persistence** | One-shot | Session-persistent |\n| **Speed** | Fast (low overhead) | Slower (9+ API calls) |\n| **Token Cost** | 1x | ~5x |\n| **Best For** | Single file analysis, search | Complex features, multi-agent collaboration |\n\n## Lifelong Learning\n\nArtibot continuously improves its routing accuracy through a session-based learning pipeline:\n\n```\nSession Start Session Active Session End\n | | |\n v v v\nLoad thresholds Record experiences Batch learning (RLVR)\nLoad System 1 cache (routing decisions + Knowledge transfer\n outcomes) Persist updated state\n```\n\n> **Note (2026-06 lean redesign):** The GRPO policy optimizer was removed (no measurable improvement). Learning now uses **verifiable-reward (RLVR)** signals (test-pass / typecheck / no-revisit) to adjust routing thresholds and promote/demote patterns between System 1 and System 2 caches. No external reward model or RL policy optimizer is used.\n\n### Knowledge Transfer\n\n| Direction | Condition | Action |\n|-----------|-----------|--------|\n| **Promote** (S2 -> S1) | 3 consecutive System 2 successes | Cache pattern in System 1 for fast retrieval |\n| **Demote** (S1 -> S2) | 2 consecutive System 1 failures | Remove from System 1, flag for System 2 re-analysis |\n\n### Learning Storage\n\n```\n~/.claude/artibot-learning/\n +-- experiences.jsonl # Raw experience log (append-only)\n +-- system1-cache.json # Promoted fast patterns\n +-- system2-cache.json # Complex pattern registry\n +-- thresholds.json # Adaptive threshold state\n +-- transfer-log.json # Promotion/demotion history\n```\n\n### Agent Teams API Tools\n\n| Tool | Purpose |\n|------|---------|\n| `Agent(type, name)` | Spawn a named teammate into the session's implicit team |\n| `TaskCreate` | Add work items to shared task list |\n| `TaskUpdate` | Set status, owner, dependencies (blockedBy/blocks) |\n| `TaskList` / `TaskGet` | View and read tasks |\n| `SendMessage` | DM, broadcast, shutdown request/response, plan approval |\n| `SendMessage(shutdown_request)` | Shut a teammate down; this IS the cleanup |\n\n### Team Levels\n\n| Level | Mode | Agents | When |\n|-------|------|--------|------|\n| **Solo** | Sub-Agent | 0 | Single file edit, quick fix |\n| **Squad** | Agent Team | 2-4 | Feature implementation, bugfix, refactoring |\n| **Platoon** | Agent Team | 5+ | Large feature, architecture change, security audit, marketing campaign |\n\n### Playbooks\n\n8 DAG playbooks (4 dev + 4 marketing) with parallel node execution.\n\n\nView all 8 playbooks\n\n**Feature:**\n```\n[Leader] plan -> [Council] design -> [Swarm] implement -> [Council] review -> [Leader] merge\n```\n\n**Bugfix:**\n```\n[Leader] analyze -> [Pipeline] fix -> [Council] verify\n```\n\n**Refactor:**\n```\n[Council] assess -> [Pipeline] refactor -> [Swarm] test -> [Council] review\n```\n\n**Security:**\n```\n[Leader] scan -> [Council] assess -> [Pipeline] fix -> [Council] verify\n```\n\n**Marketing Campaign:**\n```\n[Leader] strategy -> [Council] plan -> [Swarm] create -> [Council] review -> [Leader] launch\n```\n\n**Marketing Audit:**\n```\n[Leader] scan -> [Council] assess -> [Pipeline] optimize -> [Council] verify\n```\n\n**Content Launch:**\n```\n[Leader] plan -> [Swarm] create -> [Council] review -> [Leader] publish\n```\n\n**Competitive Analysis:**\n```\n[Council] research -> [Swarm] analyze -> [Council] synthesize -> [Leader] report\n```\n\n\n\n## Agents\n\n30 specialized agents: 1 orchestrator (CTO) + 29 specialist teammates. Fable 33% (10/30), Opus 67% (20/30) — security-reviewer stays opus via denylist.\n\n\nOrchestrator (Team Leader / CTO)\n\n| Agent | Model | Role | Team API Tools |\n|-------|-------|------|----------------|\n| **orchestrator** | fable | CTO-level team leader. Coordination only (delegation mode). | Agent(), SendMessage, TaskCreate, TaskUpdate, TaskList, TaskGet |\n\nThe orchestrator **never writes code directly**. It assembles the team, distributes tasks, coordinates between teammates, and synthesizes results.\n\n\n\n\nSpecialist Agents (29 Teammates)\n\nAll teammates have their specialist tools + team collaboration tools (`SendMessage`, `TaskList`, `TaskGet`, `TaskUpdate`).\n\n**Design & Analysis:**\n\n| Agent | Model | Specialty |\n|-------|-------|-----------|\n| architect | fable | System architecture, ADR, trade-off analysis |\n| planner | fable | Implementation planning, risk assessment |\n| llm-architect | fable | LLM architecture, prompt design, RAG |\n| investigator | fable | Evidence measurement, consistency cross-check, graded verdicts |\n| auditor | fable | Post-hoc claim census and refutation (`claim_audit` output) |\n\n**Quality & Security:**\n\n| Agent | Model | Specialty |\n|-------|-------|-----------|\n| code-reviewer | fable | Code review orchestrator (spec + quality, 2 stages) |\n| spec-reviewer | fable | Spec compliance, scope creep and over-implementation |\n| quality-reviewer | fable | Code quality, pattern adherence, coverage |\n| security-reviewer | opus | OWASP Top 10, threat modeling (pinned opus via denylist) |\n| tdd-guide | opus | TDD (RED->GREEN->REFACTOR), 80%+ coverage |\n| e2e-runner | opus | Playwright E2E testing |\n\n**Development:**\n\n| Agent | Model | Specialty |\n|-------|-------|-----------|\n| frontend-developer | opus | UI/UX, WCAG accessibility, Core Web Vitals |\n| backend-developer | opus | API, database, services |\n| database-reviewer | opus | SQL optimization, schema design |\n| typescript-pro | opus | Advanced types, strict mode |\n| build-error-resolver | opus | Build error diagnosis and auto-fix |\n\n**Utilities:**\n\n| Agent | Model | Specialty |\n|-------|-------|-----------|\n| refactor-cleaner | opus | Dead code removal, refactoring |\n| doc-updater | opus | Documentation sync, changelog |\n| content-marketer | opus | Blog, SEO, social media |\n| devops-engineer | opus | CI/CD, Docker, monitoring |\n| mcp-developer | opus | MCP server development |\n\n**Marketing:**\n\n| Agent | Model | Specialty |\n|-------|-------|-----------|\n| marketing-strategist | opus | Campaign strategy, market positioning, brand architecture |\n| data-analyst | opus | Marketing analytics, attribution modeling, KPI dashboards |\n| presentation-designer | opus | Pitch decks, marketing collateral, visual storytelling |\n| seo-specialist | opus | Technical SEO, keyword strategy, SERP optimization |\n| cro-specialist | opus | Conversion optimization, A/B testing, funnel analysis |\n| ad-specialist | opus | PPC campaigns, ad creative, ROAS optimization |\n| repo-benchmarker | fable | Repository analysis, competitive benchmarking |\n\n**Performance & Infrastructure:**\n\n| Agent | Model | Specialty |\n|-------|-------|-----------|\n| performance-engineer | opus | Performance profiling, bottleneck analysis |\n\n\n\n## Commands\n\n70+ slash commands across development, marketing, and workflow automation. Key commands: `/sc` (smart router), `/implement`, `/team`, `/code-review`, `/save`/`/resume`.\n\n\nDevelopment Commands\n\n| Command | Description |\n|---------|-------------|\n| `/sc [request]` | Smart router entry point with auto-routing |\n| `/build [target]` | Project builder with framework auto-detection |\n| `/build-fix` | Build error auto-diagnosis and fix |\n| `/implement [feature]` | Feature implementation pipeline |\n| `/improve [target]` | Evidence-based code enhancement |\n| `/design [domain]` | System design and architecture |\n\n\n\n\nAnalysis and Debugging Commands\n\n| Command | Description |\n|---------|-------------|\n| `/analyze [target]` | Multi-dimensional code/system analysis |\n| `/troubleshoot [symptoms]` | Root cause analysis |\n| `/explain [topic]` | Educational explanations |\n\n\n\n\nQuality Commands\n\n| Command | Description |\n|---------|-------------|\n| `/code-review [target]` | Code review (CRITICAL/HIGH/MEDIUM/LOW) |\n| `/test [type]` | Test runner with auto-detection |\n| `/tdd [feature]` | TDD workflow (RED->GREEN->REFACTOR) |\n| `/verify` | Validation pipeline (lint->type->test->build) |\n| `/refactor-clean [target]` | Refactoring and dead code removal |\n\n\n\n\nTeam Orchestration Commands\n\n| Command | Description |\n|---------|-------------|\n| `/orchestrate [workflow]` | Agent Teams multi-agent workflow |\n| `/spawn [mode]` | Team spawn with parallel task execution |\n| `/team [task]` | Parallel team orchestration with cross-check |\n\n\n\n\nWorkflow Commands\n\n| Command | Description |\n|---------|-------------|\n| `/plan [feature]` | Implementation planning |\n| `/task [operation]` | Task management (CRUD) |\n| `/git [operation]` | Git workflow automation |\n| `/checkpoint` | State snapshot save/restore |\n| `/daily` | Daily work recap and retrospective |\n| `/save` | Session handoff save (context restore in next session) |\n| `/resume` | Restore previous session handoff |\n| `/visual-check [url]` | Visual validation with SSIM screenshot comparison |\n| `/sc playbook list` | Browse and discover playbooks |\n\n\n\n\nDocumentation and Content Commands\n\n| Command | Description |\n|---------|-------------|\n| `/document [target]` | Documentation generation |\n| `/content [type]` | Content marketing and SEO |\n| `/learn [pattern]` | Pattern extraction and memory storage |\n\n\n\n\nMarketing Commands\n\n| Command | Description |\n|---------|-------------|\n| `/campaign [name]` | Marketing campaign orchestration and management |\n| `/seo [target]` | SEO audit, keyword research, and optimization |\n| `/social-media [platform]` | Social media content strategy and scheduling |\n| `/email-marketing [campaign]` | Email campaign design and automation |\n| `/competitive [target]` | Competitive analysis and market intelligence |\n| `/ad-campaign [platform]` | Paid advertising campaign management |\n| `/content-calendar [period]` | Editorial calendar planning and management |\n| `/ab-test [target]` | A/B testing design and analysis |\n| `/analytics [report]` | Marketing analytics and performance reporting |\n| `/funnel [stage]` | Conversion funnel analysis and optimization |\n\n\n\n\nUtility Commands\n\n| Command | Description |\n|---------|-------------|\n| `/cleanup [target]` | Technical debt reduction |\n| `/estimate [target]` | Evidence-based estimation |\n| `/index [query]` | Command catalog search |\n| `/load [path]` | Project context loading |\n| `/doctor` | Automated health check |\n| `/export [platform]` | Cross-platform agent/skill export |\n\n\n\n## Skills\n\n114 auto-activating domain skills organized in six categories.\n\n\nCore Skills (8)\n\norchestration, cognitive-routing, lifelong-learning, token-efficiency, principles, coding-standards, security-standards, testing-standards\n\n\n\n\nPersona Skills (11)\n\narchitect, frontend, backend, security, analyzer, performance, qa, refactorer, devops, mentor, scribe\n\n\n\n\nUtility Skills (8)\n\ngit-workflow, tdd-workflow, delegation, mcp-context7, mcp-playwright, mcp-coordination, continuous-learning, strategic-compact\n\n\n\n\nLanguage Skills (16)\n\nTypeScript, JavaScript, Python, Go, Rust, Java, Kotlin, Swift, C++, C#, Ruby, PHP, Scala, Elixir, R, Flutter/Dart\n\n\n\n\nMarketing Skills (34)\n\nmarketing-strategy, campaign-planning, seo-strategy, technical-seo, content-seo, social-media, email-marketing, competitive-intelligence, advertising, ab-testing, brand-guidelines, copywriting, customer-journey, data-analysis, data-visualization, lead-management, marketing-analytics, presentation-design, report-generation, segmentation, cro-page, cro-funnel, cro-forms, and more\n\n\n\n\nWorkflow Skills (7)\n\ndaily (work recap/retrospective), team (parallel orchestration), session-worklog (auto session tracking), vibe-coding (natural language coding protocol), repo-benchmarking (external repo analysis and comparison), git-worktree (worktree isolation), dynamic-context-injection (runtime context management)\n\n\n\n## Hooks\n\n27 hook registrations across 16 event types.\n\n\nHook Event Table\n\n| Event | Script | Purpose |\n|-------|--------|---------|\n| SessionStart | `session-start.js` | Environment detection, config loading |\n| PreToolUse (Write) | `pre-write.js` | Block writes to sensitive files (.env, .pem) |\n| PreToolUse (Bash) | `pre-bash.js` | Block dangerous commands (rm -rf, force push) |\n| PreToolUse (Bash) | `bash-risk-guard.js` | Classify command risk (classifyRisk): block danger, warn caution, feed autopilot pause |\n| PostToolUse (Edit) | `post-edit-format.js` | Auto-format suggestion for JS/TS |\n| PostToolUse (Bash) | `post-bash.js` | Auto-detect PR URLs after git push |\n| PostToolUseFailure | `post-tool-failure-advisor.js` | Inject corrective context for recognised failure patterns (allowlist; silent otherwise) |\n| PreCompact | `pre-compact.js` | State snapshot before context compression |\n| Stop (legacy stub) | `check-console-log.js` | No-op stub kept so sessions that cached the pre-v4.7.2 hooks.json still resolve the path; real console.log detection lives in `dev-verify-gate.js` |\n| UserPromptSubmit | `user-prompt-handler.js` | Intent detection and agent suggestion |\n| UserPromptSubmit | `runtime-prompt.js` | Second-stage prompt enrichment — builds the runtime envelope (replaced the former cognitive-router hook) |\n| SubagentStart/Stop | `subagent-handler.js` | Teammate registration/deregistration tracking |\n| TeammateIdle | `team-idle-handler.js` | Alert idle teammates about pending tasks |\n| SessionEnd | `session-end.js` | Persist session state |\n| SessionEnd | `memory-tracker.js` | Summarize the session and persist learnings |\n| SessionEnd | `http-notify.js` | HTTP webhook notifications (Slack/Discord/generic) |\n\n\n\n## Auto-Update\n\nArtibot checks for new versions on session start via GitHub Releases API (24h cached).\n\n```\nSession start:\n Artibot v1.8.0 initialized\n ✅ You are running the latest version\n```\n\n**`/update` command** (flat install; `/artibot:update` if installed via marketplace)**:**\n\n| Flag | Behavior |\n|------|----------|\n| `--check` | Check version only (default) |\n| `--force` | Clear cache and force reinstall |\n| `--dry-run` | Show plan without executing |\n\n## MCP Integration\n\nArtibot integrates with MCP servers for extended capabilities:\n\n**Context7** -- Library and framework documentation lookup\n```json\n{\n \"context7\": {\n \"command\": \"npx\",\n \"args\": [\"-y\", \"@upstash/context7-mcp@latest\"]\n }\n}\n```\n\n**Playwright** -- Cross-browser E2E testing, performance metrics, visual testing\n```json\n{\n \"playwright\": {\n \"command\": \"npx\",\n \"args\": [\"-y\", \"@executeautomation/playwright-mcp-server\"]\n }\n}\n```\n\n## Cross-Platform Compatibility\n\nArtibot includes platform adapters for use beyond Claude Code:\n\n| Platform | Adapter | Status |\n|----------|---------|--------|\n| **Claude Code** | Native (no adapter needed) | Full support |\n| **Gemini CLI** | `adapters/gemini-cli.js` | Agent mapping, command translation |\n| **OpenAI Codex** | `adapters/openai-codex.js` | Tool mapping, prompt adaptation |\n| **Cursor** | `adapters/cursor.js` | Extension integration, command bridging |\n\nEach adapter translates Artibot's Agent Teams API calls into the target platform's native orchestration primitives while preserving the CTO-led coordination pattern.\n\n## Plugin Structure\n\n\nView directory tree\n\n```\nplugins/artibot/\n+-- .claude-plugin/\n| +-- plugin.json # Plugin manifest\n+-- agents/ # 30 agent definitions\n| +-- orchestrator.md # CTO / Team leader (Agent Teams API)\n| +-- [17 dev specialists].md # Development teammates\n| +-- [8 marketing agents].md # Marketing specialists\n+-- commands/ # 79 slash commands\n| +-- sc.md # Smart router\n| +-- daily.md # Daily work recap and retrospective\n| +-- team.md # Parallel team orchestration\n| +-- orchestrate.md # Team orchestration (named Agent spawns)\n| +-- spawn.md # Team spawn (parallel execution)\n| +-- [25 dev commands].md\n| +-- [10 marketing commands].md\n+-- skills/ # 114 skill directories\n| +-- orchestration/ # Delegation mode + team routing\n| +-- cognitive-routing/ # System 1/2 dual-process routing\n| +-- lifelong-learning/ # GRPO batch learning + knowledge transfer\n| +-- delegation/ # Sub-Agent/Team strategies\n| +-- [23 dev skills]/\n| +-- [23 marketing skills]/\n| +-- [7 workflow skills]/\n+-- rules/ # 10 auto-activating rules\n| +-- dev-protocol.md # DEV (Decompose-Execute-Verify) protocol\n| +-- quality-gates.md # Quality enforcement gates\n| +-- agent-coordination.md # Agent collaboration patterns\n| +-- [7 domain rules].md\n+-- hooks/\n| +-- hooks.json # Hook event mappings\n+-- scripts/\n| +-- hooks/ # 70 hook scripts (ESM)\n| +-- ci/ # 20 CI scripts\n| +-- evals/ # Runtime eval suite\n| +-- utils/\n+-- lib/\n| +-- core/ # Core modules (platform, config, cache, playbook-parser, playbook-registry, guard-registry, event-bus, blocked-patterns)\n| +-- runtime/ # Runtime pipeline (15 middleware modules; 11-stage default chain: lifecycle, router, memory, skills, tasks, subagents, guardrail, summarization, token-usage, checkpoint, cache-roi)\n| +-- visual/ # Visual validation (SSIM differ, style-fixer, validator)\n| +-- intent/ # Intent detection (language, trigger)\n| +-- context/ # Context management (session)\n| +-- privacy/ # PII protection (pii-detector, pii-scrubber, homoglyph, token-rotation, differential-privacy)\n| +-- learning/ # Lifelong learning (39 modules: memory, GRPO, pattern-analyzer, tool-history, rule-extractor, skill-injector, knowledge-demotion, etc.)\n| +-- adapters/ # Cross-platform adapters\n+-- output-styles/ # 7 output styles (default, compressed, intelligence, mentor, narrative, report, team-dashboard) + tokens.md\n+-- templates/ # 5 writing templates\n+-- artibot.config.json # Plugin config (Agent Teams settings)\n+-- package.json # Node.js ESM runtime\n+-- .mcp.json # MCP server configuration\n```\n\n\n\n## Configuration\n\nKey settings in `artibot.config.json`:\n\n| Setting | Description | Default |\n|---------|-------------|---------|\n| `version` | Plugin version | `1.14.1` |\n| `cognitive.router.threshold` | System 1/2 boundary | `0.4` |\n| `cognitive.router.adaptRate` | Per-feedback adjustment step | `0.05` |\n| `cognitive.system1.maxLatency` | System 1 max response time (ms) (unused — engine removed) | `100` |\n| `cognitive.system1.minConfidence` | System 1 minimum confidence (unused — engine removed) | `0.6` |\n| `cognitive.system2.maxRetries` | System 2 max retry attempts (unused — engine removed) | `3` |\n| `cognitive.system2.sandboxEnabled` | Enable sandbox for high-risk ops (unused — engine removed) | `true` |\n| `learning.lifelong.batchSize` | Experiences per lifelong-learning batch | `50` |\n| `learning.lifelong.grpoGroupSize` | Experiences per comparison group | `5` |\n| `learning.knowledgeTransfer.promotionThreshold` | Consecutive successes to promote | `3` |\n| `learning.knowledgeTransfer.demotionThreshold` | Consecutive failures to demote | `2` |\n| `team.engine` | Team engine | `claude-agent-teams` |\n| `team.delegationMode` | Leader coordination-only mode | `true` |\n| `team.maxTeammates` | Max concurrent teammates | `null` (unlimited) |\n| `team.ctoAgent` | CTO agent name | `orchestrator` |\n| `team.levels.squad` | Squad team size | `2-4` |\n| `team.levels.platoon` | Platoon team size | `5+` |\n| `automation.intentDetection` | Auto intent detection | `true` |\n| `automation.supportedLanguages` | Supported languages | `en, ko, ja, zh` |\n\n## Validation\n\n```bash\nnode scripts/validate.js # Full validation\nnode scripts/ci/validate-agents.js # Agent validation\nnode scripts/ci/validate-skills.js # Skill validation\nnode scripts/ci/validate-commands.js # Command validation\nnode scripts/ci/validate-hooks.js # Hook validation\n```\n\n## Best Practices\n\n- **Use `/sc` for auto-routing** -- Let the smart router pick the optimal command and delegation mode\n- **Trust the delegation scoring** -- Simple tasks use fast sub-agents; complex tasks get full teams\n- **Enable Agent Teams** -- Without `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`, team features won't work\n- **Let teammates self-claim** -- The Swarm pattern works best when teammates pick tasks from the shared list\n- **Use playbooks for common workflows** -- 8 playbooks (4 dev + 4 marketing) provide proven team patterns\n- **Leverage marketing specialists** -- Use `/campaign` and related commands for end-to-end marketing orchestration\n\n## When to Use Artibot\n\n**Use for:**\n- Complex feature implementation spanning multiple files and domains\n- Security audits requiring multi-perspective analysis\n- Large-scale refactoring with coordinated testing\n- Architecture decisions needing specialist input\n- Marketing campaigns requiring strategy, content, SEO, and analytics coordination\n- Competitive analysis with data-driven insights\n- Any task benefiting from parallel agent collaboration\n\n**Don't use for:**\n- Single-line bug fixes\n- Trivial edits or formatting changes\n- Tasks where a single agent is sufficient\n- Quick questions or explanations (unless you want team discussion)\n\n## Troubleshooting\n\n### Agent Teams not working\n\n**Issue**: Agent or SendMessage not available\n\n**Solution**: Ensure the environment variable is set:\n```json\n// ~/.claude/settings.json\n{\n \"env\": {\n \"CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS\": \"1\"\n }\n}\n```\n\n### Teammates not picking up tasks\n\n**Issue**: Tasks stay in \"pending\" status\n\n**Solution**:\n- Verify tasks have no unresolved `blockedBy` dependencies\n- Check that teammates are spawned with a `name` (an unnamed teammate cannot be messaged)\n- Use `TaskUpdate` to explicitly assign if self-claim isn't working\n\n### High token usage\n\n**Issue**: Team orchestration consuming too many tokens\n\n**Solution**:\n- Use Sub-Agent mode for simpler tasks (score < 0.4)\n- Reduce team size with `--agents` flag\n- Enable token efficiency with `--uc` flag\n\n## Contributing\n\nSee [CONTRIBUTING.md](CONTRIBUTING.md) for the full guide on adding skills, agents, and commands.\n\n1. Fork this repository\n2. Create a feature branch\n3. Follow the existing plugin structure and conventions\n4. Run `npm run ci` to validate\n5. Submit a pull request\n\n## Author\n\n**Artience** ([@Yoodaddy0311](https://github.com/Yoodaddy0311))\n\n## Quick Links\n\n| Resource | Path |\n|---|---|\n| **Installation guide** | [`INSTALL.md`](./INSTALL.md) |\n| **Contributing guide** | [`CONTRIBUTING.md`](./CONTRIBUTING.md) |\n| **CHANGELOG** (root summary) | [`CHANGELOG.md`](./CHANGELOG.md) |\n| **CHANGELOG** (full detail) | [`plugins/artibot/CHANGELOG.md`](./plugins/artibot/CHANGELOG.md) |\n| **artibot plugin README** (Claude Code) | [`plugins/artibot/README.md`](./plugins/artibot/README.md) |\n| **artibot-cowork plugin README** (Claude Cowork) | [`plugins/artibot-cowork/README.md`](./plugins/artibot-cowork/README.md) |\n| **MCP server usage** | `plugins/artibot/docs/MCP-SETUP.md` |\n| **Marketplace submission package** | [`plugins/artibot/_marketplace/`](./plugins/artibot/_marketplace/) |\n\n## When to use which\n\n```\n ┌──────────────────────────┐\n Developer task → │ artibot (Claude Code) │ ← TDD, /code-review, /implement, /orchestrate\n └──────────────────────────┘\n\n ┌──────────────────────────┐\n Knowledge-worker task → │ artibot-cowork (Cowork) │ ← /campaign, long-form writing, KR SEO\n └──────────────────────────┘\n```\n\nYou can install **both** in the same Anthropic account — `artibot` runs in your Claude Code terminal sessions, `artibot-cowork` runs in your Cowork chat. They never interfere because they target different runtimes.\n\n## Version\n\n**4.13.0** (2026-05-19) — feat: `/save` single-shot session handoff command (+`/resume`) — `.artibot/HANDOFF.md` produces 5s next-session context restore. Safety boost: YAML frontmatter (machineId/createdAt/branch), git-lock graceful fail (5s timeout), 10m archive throttle, atomic advisor consumed marking. See [CHANGELOG](./plugins/artibot/CHANGELOG.md) for full history.\n\n**4.12.0** (2026-05-19) — Comprehensive audit fixes: security hardening (timing-safe auth, K_SERVICE bypass removed, shell-injection surfaces eliminated), manifest drift sync, layer-cycle fix.\n\n[📖 Full release history — CHANGELOG](./plugins/artibot/CHANGELOG.md)\n\n## License\n\nBusiness Source License 1.1 — see [LICENSE](LICENSE) for the full text. Free for\nnon-production use and for production use that is not a Commercial Competing\nProduct (a paid/commercial product substantially similar to Artibot). Converts\nto Apache License 2.0 on 2030-02-20 (or the fourth anniversary of a given\nversion's first public release, whichever comes first). Both `plugins/artibot/`\nand `plugins/artibot-cowork/` ship under the same license.\n",
"bytes": 47132,
"sha": "aa414202af2495c32973d691778744ae3b81aeb7b5a91be250c2845df1b7485f",
"repo_slug": "yoodaddy0311/artibot",
"fonte": "repo",
"truncated": false,
"api": "https://agentalog.com/api/listings/plg_yoodaddy0311_artibot_artibot_0cb5a14a/readme"
}
![\"Artibot](\"assets/artibot-mascot.jpeg\")
\n