{"id":"ce9b02e8-cd69-46b8-932b-0b1222e6da48","slug":"immamdouhaboammar-agentic-workflow","kind":"skill","name":"agentic-workflow","summary":"Pluripotent stem-cell framework and universal agentic toolchain for autonomous multi-agent workflows","owner":"imMamdouhaboammar","repoUrl":"https://github.com/imMamdouhaboammar/agentic-workflow","license":"MIT","licenseTier":"permissive","trust":"community","stars":0,"match":null,"fewStars":true,"format":"skill","createdAt":"2026-09-06T08:30:24.000Z","addedAt":"2026-10-21T12:59:00.014Z","logoUrl":"https://avatars.githubusercontent.com/u/58908124?u=323b43370c2cc5fe374e75b8cf72cc4769376c0b&v=4","linkOnly":false,"collection":null,"skillPath":null,"categories":["developer-tools"],"tags":["antigravity","bun","claude-code","codex","cursor","gemini-cli","multi-agent-systems","skills-sh"],"runtime":"node","local":true,"cloudCapable":false,"cloudCapability":"unknown","cloudReadyOnPlatform":false,"needsApiKey":false,"reviewed":false,"syncedAt":"2026-10-23T05:36:09.615Z","sample":false,"url":"https://conduits.dev/tools/immamdouhaboammar-agentic-workflow","language":"Python","transports":["stdio"],"install":{"npm":"@mamdouh-aboammar/agentic-workflow"},"requirements":{"env":[],"upstreamMayCharge":false},"tools":null,"readme":"
\n\n# ⚡ AgenticWorkflow ⚡\n\n### Pluripotent Stem-Cell Framework & Universal Agentic Toolchain\n**Deterministic Quality Gates • Multi-Engine Autopilot • Single-File SOT • TOON v4.1 Density**\n\n[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg?style=flat-square)](LICENSE)\n[![Bun](https://img.shields.io/badge/Runtime-Bun%20%3E%3D1.0-FBF0DF?style=flat-square&logo=bun&logoColor=black)](https://bun.sh)\n[![Python](https://img.shields.io/badge/Python-%3E%3D3.10-3776AB?style=flat-square&logo=python&logoColor=white)](https://python.org)\n[![TypeScript](https://img.shields.io/badge/TypeScript-5.0-3178C6?style=flat-square&logo=typescript&logoColor=white)](https://www.typescriptlang.org/)\n[![npm](https://img.shields.io/badge/npm-agentic--workflow-CB3837?style=flat-square&logo=npm&logoColor=white)](https://www.npmjs.com/package/agentic-workflow)\n[![PyPI](https://img.shields.io/badge/PyPI-agentic--workflow-3775A9?style=flat-square&logo=pypi&logoColor=white)](https://pypi.org/project/agentic-workflow)\n[![Skills.sh](https://img.shields.io/badge/Skills.sh-Compatible-000000?style=flat-square&logo=vercel&logoColor=white)](https://skills.sh)\n[![Claude Code](https://img.shields.io/badge/Claude%20Code-Certified%20Skill-D97706?style=flat-square&logo=anthropic&logoColor=white)](https://claude.ai)\n[![Cursor](https://img.shields.io/badge/Cursor-Rules%20%26%20Skills-000000?style=flat-square&logo=cursor&logoColor=white)](https://cursor.com)\n[![CI](https://img.shields.io/badge/CI-Passing-10B981?style=flat-square&logo=githubactions&logoColor=white)](https://github.com/imMamdouhaboammar/agentic-workflow/actions)\n\n

\n Quickstart •\n Agent Prompt •\n Why It Exists •\n Architecture •\n SDK Usage •\n CLI Reference •\n Supportive Tools •\n License\n

\n\n
\n\n---\n\n## 🚀 Universal Installation & Quickstart\n\nAgenticWorkflow is packaged and distributable across every major package registry, AI harness, and container platform:\n\n### 1. 🤖 One-Click Agent Self-Install & Init Prompt\n\nPaste this prompt directly into your AI coding agent (Claude Code, Antigravity / Gemini CLI, Cursor, Codex, OpenCode, or Windsurf) to let it autonomously install, configure, initialize, and verify `AgenticWorkflow` in one shot:\n\n
\n👉 Click to expand the Smart Agent Prompts (Master Prompt & Quick-Fire)\n\n
\n\n> [!TIP]\n> **Copy & Paste into your Agent Chat:**\n\n````markdown\nYou are acting as an autonomous DevOps & Systems Integration Agent. \nYour objective is to install, configure, initialize, and verify the `agentic-workflow` framework in this workspace in a single uninterrupted execution loop.\n\nExecute the following 4 phases autonomously:\n\n### Phase 1: Universal Installation\n1. Detect available runtimes on the system (`python3`, `bun`, `npm`, `curl`, `git`).\n2. Run the universal one-line installer:\n ```bash\n curl -fsSL https://raw.githubusercontent.com/imMamdouhaboammar/agentic-workflow/main/install.sh | bash\n ```\n *(Fallback if offline or in a local clone: run `./install.sh` or `bun add -g agentic-workflow` / `npm install -g agentic-workflow`)*.\n3. Ensure `~/.local/bin` is in `$PATH` so the `agentic-workflow` CLI binary is immediately executable.\n\n### Phase 2: Host & Workspace Configuration\n1. Detect the current agent harness and verify registration in the appropriate skill directories:\n - Claude Code: `~/.claude/skills/agentic-workflow`\n - Gemini CLI / Antigravity: `~/.gemini/config/skills/agentic-workflow`\n - Cursor: `~/.cursor/skills/agentic-workflow`\n - Codex / OpenCode: `~/.codex/skills/agentic-workflow`\n - Universal Agent Kernel: `~/.agents/skills/agentic-workflow`\n2. If this workspace lacks an `AGENTS.md` directive, link or import the AgenticWorkflow constitution (`AGENTS.md` and `GEMINI.md` / `CLAUDE.md`) so all subagents inherit:\n - Absolute Criterion 1: Deliverable Quality over speed/token cost.\n - Absolute Criterion 2: Single-File SOT (`state.yaml`) with single-writer lock.\n - Absolute Criterion 3: Code Change Protocol (CCP: Understand Intent → Ripple Effect Analysis → Step Plan).\n - 4-Layer Quality Assurance: L0 Anti-Skip → L1 Verification → L1.5 pACS Pre-mortem scoring → L2 Adversarial Review.\n3. Provision supportive tools by running:\n ```bash\n agentic-workflow integrations install\n ```\n Confirm integration of Ponytail (YAGNI & surgical diffs), TOON v4.1 (token compression), Fable (circuit breaker), and Caveman (concise mode).\n\n### Phase 3: Initialization (`init`)\n1. Run the native initialization command:\n ```bash\n agentic-workflow init\n ```\n2. Confirm that:\n - Runtime directories (`context-snapshots/`, `autopilot-logs/`, `diagnosis-logs/`) are generated.\n - Core hook scripts in `.claude/hooks/scripts/` have validated syntax and execution permissions.\n - The skills mesh index (`core/skills_indexer.py`) is indexed and queryable.\n\n### Phase 4: Validation & Status Report\n1. Verify the setup by running:\n ```bash\n agentic-workflow validate\n agentic-workflow integrations status\n ```\n2. Print a concise completion summary formatted in TOON v4.1 or a clean Markdown table showing CLI status, host skill registration paths, supportive tools readiness, and quality gates status.\n\nProceed and execute all steps now without stopping for manual input.\n````\n\n#### ⚡ Quick-Fire Version\n```text\nRun the universal installer for AgenticWorkflow (`curl -fsSL https://raw.githubusercontent.com/imMamdouhaboammar/agentic-workflow/main/install.sh | bash`), link CLI to PATH, run `agentic-workflow init` to configure SOT runtime and supportive tools (Ponytail, TOON, Fable, Caveman), and run `agentic-workflow validate` to confirm 100% readiness. Report the final status table when done.\n```\n\n
\n\n### 2. Agent Skill Hubs (Zero-Install Agent Registration)\n\n```bash\n# Skills.sh / Vercel Ecosystem (Any Agent)\nnpx skills add imMamdouhaboammar/agentic-workflow\n\n# Universal One-Line Installer (Claude, Gemini, Cursor, Codex, OpenCode)\ncurl -fsSL https://raw.githubusercontent.com/imMamdouhaboammar/agentic-workflow/main/install.sh | bash\n```\n\n### 3. Package Managers (CLI & SDK)\n\n| Registry / Host | Command | Usage |\n|---|---|---|\n| **Bun (Instant CLI)** | `bunx @mamdouh-aboammar/agentic-workflow [command]` | Zero-install CLI execution |\n| **Bun (Library)** | `bun add @mamdouh-aboammar/agentic-workflow` | TypeScript / Bun SDK dependency |\n| **npm / npx (Node)** | `npx @mamdouh-aboammar/agentic-workflow [command]` | Zero-install Node CLI execution |\n| **npm (Library)** | `npm install @mamdouh-aboammar/agentic-workflow` | Node.js ESM library dependency |\n| **PyPI (Python)** | `pip install agentic-workflow-toolchain` | Python library & console script |\n| **Homebrew (macOS/Linux)** | `brew install imMamdouhaboammar/tap/agentic-workflow` | System binary via Homebrew |\n| **Docker Container** | `docker run -it ghcr.io/immamdouhaboammar/agentic-workflow` | Isolated, containerized runner |\n\n---\n\n## ⚡ Why AgenticWorkflow Exists\n\nMost AI workflows fail in production due to three compounding traps:\n1. **Hallucinated Progress**: Agents mark tasks complete without verifying actual deliverables on disk.\n2. **Context Amnesia**: Sessions reset or compact, losing critical context and historical failures.\n3. **Unchecked Drift**: Multi-agent swarms mutate shared state simultaneously, causing race conditions and logic divergence.\n\nAgenticWorkflow eliminates these failure modes with a 2-stage execution model backed by deterministic Python and TypeScript safety rails:\n\n```mermaid\nflowchart LR\n Phase1[\"Phase 1: Workflow Design (workflow.md blueprint)\"] --> Phase2[\"Phase 2: Workflow Implementation (Executing Autonomous System)\"]\n```\n\nCreating `workflow.md` is only half the journey. **The ultimate goal is that the workflow executes reliably and produces verified deliverables.**\n\n---\n\n## 🏛️ 3-Stage Core Architecture\n\nEvery workflow strictly follows three sequential stages:\n\n```mermaid\ngraph TD\n subgraph ResearchStage [\"1. Research Stage\"]\n R1[\"Information Gathering\"] --> R2[\"Domain Analysis & Fact Verification\"]\n end\n\n subgraph PlanningStage [\"2. Planning Stage\"]\n P1[\"State Formulation (state.yaml SOT)\"] --> P2[\"Human / Autopilot Review & Approval\"]\n end\n\n subgraph ImplementationStage [\"3. Implementation Stage\"]\n I1[\"Autonomous Execution & Tool Orchestration\"] --> I2[\"4-Layer Quality Gates & Final Deliverables\"]\n end\n\n ResearchStage --> PlanningStage\n PlanningStage --> ImplementationStage\n```\n\n1. **Research** — Information gathering, competitive benchmarking, and deep domain analysis.\n2. **Planning** — Architecture blueprint formulation, task decomposition, and human/autopilot sign-off.\n3. **Implementation** — Multi-agent tool execution, code generation, and artifact verification.\n\n---\n\n## 🛡️ 4-Layer Quality Assurance Stack\n\nEvery step completion must pass up to 4 verification layers before the Orchestrator advances the Single Source of Truth (`state.yaml`):\n\n```mermaid\nflowchart TD\n StepRun[\"Agent Executes Step Task\"] --> L0[\"L0: Anti-Skip Physical Guard (File exists & >= 100 bytes)\"]\n L0 -->|\"PASS\"| L1[\"L1: Verification Gate (100% functional goal achievement)\"]\n L0 -->|\"FAIL\"| Retry[\"Deterministic Retry / Diagnosis\"]\n L1 -->|\"PASS\"| L15[\"L1.5: pACS Self-Rating (F/C/L Pre-mortem scoring)\"]\n L1 -->|\"FAIL\"| Retry\n L15 -->|\"RED: <50\"| Retry\n L15 -->|\"GREEN / YELLOW\"| L2[\"L2: Adversarial Review (@reviewer + @fact-checker)\"]\n L2 -->|\"PASS\"| SOTUpdate[\"Update SOT state.yaml (current_step + 1)\"]\n L2 -->|\"FAIL\"| AbductiveDiag[\"Abductive Diagnosis (diagnose_context.py)\"]\n AbductiveDiag --> Retry\n```\n\n| Layer | Gate Name | Target Verified | Mechanism |\n|---|---|---|---|\n| **L0** | Anti-Skip Guard | Physical deliverable exists and size $\\ge 100$ bytes | Deterministic Python hook |\n| **L1** | Verification Gate | 100% achievement of declared task acceptance criteria | Semantic agent self-verification |\n| **L1.5** | pACS Calibration | 3D confidence scoring (Faithfulness, Completeness, Logic) | Pre-mortem protocol ($\\min(F, C, L)$) |\n| **L2** | Adversarial Review | Independent critique, claim audit, and web fact-checking | `@reviewer` + `@fact-checker` subagents |\n\n---\n\n## 💻 Dual-Language SDK Usage\n\n### TypeScript & Bun (`npm install agentic-workflow` or `bun add agentic-workflow`)\n\n```typescript\nimport { \n AutopilotEngine, \n HookDispatcher, \n IntegrationInstaller, \n encodeToon, \n calculateTokenSavings \n} from 'agentic-workflow';\n\n// 1. Token-Oriented Object Notation (v4.1) compression\nconst data = {\n users: [\n { id: 1, name: \"Alice\", role: \"architect\" },\n { id: 2, name: \"Bob\", role: \"reviewer\" }\n ]\n};\nconst toonData = encodeToon(data);\nconsole.log(`Compressed TOON:\\n${toonData}`);\n\n// 2. Hook Dispatcher evaluation\nconst dispatcher = new HookDispatcher(process.cwd());\nconst check = dispatcher.dispatch({\n event_id: \"evt_1\",\n source: \"cli\",\n hook_type: \"pre_command\",\n timestamp: Date.now(),\n command: \"git status\"\n});\nconsole.log(`Hook verdict: ${check.verdict}`);\n```\n\n### Python (`pip install agentic-workflow`)\n\n```python\nfrom agentic_workflow import (\n AutopilotEngine, \n HookDispatcher, \n IntegrationInstaller,\n CleanCodeChecker,\n MultiAgentManager\n)\n\n# 1. Launch Autopilot Engine\nengine = AutopilotEngine(project_dir=\".\", auto_approve=True)\nengine.plan_default_workflow(\n title=\"Data Ingestion Pipeline\", \n goal=\"Autonomous end-to-end data ingestion with quality gates\"\n)\nsuccess = engine.run_all()\n\n# 2. Check Supportive Tools Status\ninstaller = IntegrationInstaller(project_dir=\".\")\nresults = installer.check_all()\nfor r in results:\n print(f\"- {r.name}: {r.status}\")\n```\n\n---\n\n## ⚙️ CLI Reference\n\n```bash\n# Launch autonomous end-to-end autopilot workflow with self-fueling & energy management\nagentic-workflow autopilot --title \"Production Pipeline\" --goal \"Autonomous Delivery\"\n\n# Run Clean Code Guard audit pass (SOLID, 24 Imperatives, AI failure modes)\nagentic-workflow guard [directory]\n\n# Execute AI Engineer fairness, drift, and prompt-injection evaluation gates\nagentic-workflow eval\n\n# Query multi-agent observable trace logs and spans\nagentic-workflow traces\n\n# Manage supportive tools (Ponytail, TOON, Fable, Caveman) & lifecycle\nagentic-workflow integrations status\nagentic-workflow integrations install\nagentic-workflow integrations phase planning\n\n# Token-Oriented Object Notation (v4.1) benchmarks and conversion\nagentic-workflow toon benchmark\nagentic-workflow toon convert \n\n# Initialize infrastructure, SOT runtime directories, and supportive tools\nagentic-workflow init\n\n# Validate workflow.md, SOT schema, and pACS integrity\nagentic-workflow validate\n\n# Check current workflow progress and observability dashboard\nagentic-workflow status\n\n# Run full automated test suite (16 suites: safety, guard, MAS, engines, integrations)\nagentic-workflow test\n```\n\n---\n\n## 🧰 Supportive Tools Ecosystem\n\nAgenticWorkflow automatically provisions and directs specialized supportive tools across its execution phases without manual user overhead:\n\n| Supportive Tool | Role & Category | Designated Lifecycle Phase |\n|---|---|---|\n| **[Ponytail](https://github.com/DietrichGebert/ponytail)** | **Simplicity Governor & Anti-Debt** | **Planning & Implementation**: Enforces YAGNI ladder, stdlib-first, and shortest working surgical diffs. |\n| **[TOON](https://github.com/toon-format/toon)** | **Token-Oriented Object Notation (v4.1)** | **Continuous Data Protocol**: Cuts structured data and state tokens by 30-60% across all deliverables and logs. |\n| **[Fable](https://github.com/imMamdouhaboammar/get-fable)** | **Lifecycle Harness & Continuation** | **Execution & Handoff**: Arms circuit breakers (halts on failure streak $\\ge 2$) and generates durable continuation state (`.fable/`). |\n| **[Caveman](https://github.com/JuliusBrussee/caveman)** | **Terse Communication Mode** | **Continuous Protocol**: Strips conversational fluff to cut output tokens by 65-75% while keeping code and errors exact. |\n\n---\n\n## 📜 Absolute Criteria (Canon)\n\nThese constitutional rules govern every design, execution, and modification decision:\n\n1. **Absolute Criterion 1: Quality of the Final Deliverable**\n > Speed, token cost, workload, and length limits are completely ignored. The sole criterion for every decision is the **quality of the final deliverable**.\n2. **Absolute Criterion 2: Single-File SOT + Hierarchical Memory**\n > All shared workflow state is concentrated in a single file (`state.yaml`). Write permission belongs exclusively to the Orchestrator / Team Lead. Parallel agents never mutate shared files simultaneously.\n3. **Absolute Criterion 3: Code Change Protocol (CCP)**\n > Before writing, modifying, adding, or deleting code, you must perform **Step 1 (Understand Intent) → Step 2 (Ripple Effect Analysis) → Step 3 (Change Plan)**. Governed by Coding Anchor Points (CAP-1~4).\n\n---\n\n## 📖 Documentation Roadmap\n\n1. **README.md** (This document) — High-level bird's-eye overview and distribution hub.\n2. [`soul.md`](soul.md) — The philosophical core and DNA inheritance principles.\n3. [`AGENTICWORKFLOW-ARCHITECTURE-AND-PHILOSOPHY.md`](AGENTICWORKFLOW-ARCHITECTURE-AND-PHILOSOPHY.md) — Architectural design and theoretical foundations.\n4. [`DECISION-LOG.md`](DECISION-LOG.md) — Complete historical record of architectural decisions (ADRs).\n5. [`AGENTICWORKFLOW-USER-MANUAL.md`](AGENTICWORKFLOW-USER-MANUAL.md) — Practical step-by-step operating instructions.\n6. [`AGENTS.md`](AGENTS.md) — Universal directive and constitutional rules.\n7. [`docs/protocols/`](docs/protocols/) — Deep-dive execution protocols.\n\n---\n\n## 📄 License\n\nMIT License © 2026 Mamdouh Aboammar & Yoonsik Choi. All rights reserved.\n"}