AI Skill Engineer

Transform a single human idea into a complete, production-ready application — autonomously.

🏆 Orchestrate 24 AI engineering roles | 10-phase autonomous workflow | Plug any LLM provider

---

Why AI Skill Engineer?

Unlike code assistants that generate snippets, AI Skill Engineer runs a complete multi-skill, multi-phase engineering workflow — from idea to production-ready delivery package.

Idea → Understand → Plan → Discover Skills → Build → Review → Fix → Validate → Approve → Optimize → Deliver

What makes it different
24 simulated expert roles	Product strategists, solution architects, frontend/backend/AI engineers, security, QA, DevOps, and more
DAG-based parallel execution	Skills run in dependency-respecting parallel groups
State persistence	Resume from the last completed phase — no data loss
Validation pipeline	Type-checking, linting, security scanning, performance tests built in
Pluggable executors	Connect any LLM provider (OpenAI, Anthropic, Ollama, etc.) via a simple interface
Delivery packaging	Complete project with architecture docs, code, tests, deployment, and runbooks

---

Quick Start

git clone https://github.com/forgephantom/Ai-skill-engineer.git
cd Ai-skill-engineer && npm install && npm run build && npm link
ai-se init "Build a SaaS platform for compliance reporting"
ai-se run --project <project-id>

📖 Full walkthrough with sample output →

---

Step-by-Step Guide

1. Installation

Prerequisites, cloning, dependency installation, building, and linking the CLI.

Example:

git clone https://github.com/forgephantom/Ai-skill-engineer.git
cd Ai-skill-engineer
npm install
npm run build
npm link
ai-se doctor    # Verify everything works

2. Initialize a Project

Turn your idea into a structured project with a single command.

Example:

ai-se init "A marketplace connecting freelance developers with non-technical founders" \
  --name freelancer-marketplace \
  --output ./projects \
  --no-human-approval

3. Run the Workflow

Execute all 10 phases autonomously.

Example:

ai-se run --project proj-a1b2c3d4 --no-human-approval

4. Monitor & Resume

Check status or resume an interrupted workflow.

Example:

ai-se status --project proj-a1b2c3d4
ai-se resume --project proj-a1b2c3d4

---

End-to-End Example

See the complete Freelancer Marketplace use case — a full 10-phase walkthrough with detailed artifacts, code samples, and terminal output:

📄 docs/use-cases/freelancer-marketplace.md

Quick start the example:

ai-se init "A marketplace connecting freelance developers with non-technical founders" \
  --name freelancer-marketplace \
  --output ./output \
  --no-human-approval
ai-se run --project ./output/proj-a1b2c3d4 --no-human-approval

---## 10-Phase Workflow

#	Phase	Skills	Key Outputs
1	Understand	Business Analyst, Product Strategist	Vision, goals, requirements, risks
2	Plan	Product Manager, Solution Architect, Technical Writer	PRD, user stories, tech spec, roadmap
3	Discover Skills	Skill Discovery Engine	Skill dependency DAG with parallel groups
4	Build	All project skills (parallel groups)	Architecture, code, UI, API, DB, infra
5	Review	Code Reviewer	Findings with severity breakdown
6	Fix	Code Fixer	Remediated artifacts, regression check
7	Validate	Validation Engine	Static analysis, lint, type-check, test results
8	Human Approval	Principal Engineer Simulator	Approval or feedback (auto or manual)
9	Optimize	Optimization Engine	Performance, security, cost, UX improvements
10	Deliver	Delivery Engineer	Complete package with docs, deploy guide, runbooks

Each phase is explained in detail with example output at docs/workflow-phases.md.

---

Built-in Skills (24)

product-strategist · business-analyst · product-manager · solution-architect · technical-writer · ux-designer · ui-designer · frontend-engineer · backend-engineer · mobile-engineer · ai-engineer · database-engineer · cloud-engineer · devops-engineer · security-engineer · qa-engineer · documentation-engineer · code-reviewer · code-fixer · validation-engine · optimization-engine · delivery-engineer · principal-engineer-simulator · skill-discovery-engine

---

Connect AI Providers

Skills plug into any LLM provider via a simple executor interface:

Provider	Setup
Anthropic Claude	npm install @anthropic-ai/sdk + ANTHROPIC_API_KEY
OpenAI GPT-4	npm install openai + OPENAI_API_KEY
Ollama (Local)	No SDK needed — uses fetch
Google Gemini	npm install @google/generative-ai + GEMINI_API_KEY

Full code examples for each provider at docs/skill-executors.md.

---

CLI Reference

Command	Description
init [idea]	Initialize a new project
run --project <id>	Execute the workflow
status --project <id>	Show project state
resume --project <id>	Resume interrupted workflow
stop --project <id>	Stop running project
validate --project <id>	Run validation pipeline
doctor	Check system health
skills list/show <id>	Manage skill definitions
templates list/show <name>	Manage templates
config show/set <key> <value>	Manage project config

Full details with examples at docs/cli-reference.md.

---

Configuration

Configure via project-config.json or CLI:

Option	Default	Description
model	claude-3-5-sonnet	Default AI model
maxParallelSkills	4	Max concurrent skill executions
validationLevel	standard	strict, standard, or minimal
humanApprovalRequired	true	Require human approval gate

---

Custom Skills

Create your own skills with a YAML definition file and optional executor:

id: my-skill
name: My Skill
version: "1.0.0"
mission: "One-sentence mission statement"
model: claude-3-5-sonnet
responsibilities:
  - "Do X"
  - "Do Y"
inputs:
  - artifact_id: "input-name"
    contract: "json"
    required: true
outputs:
  - artifact_id: "output-name"
    contract: "markdown"
dependencies: ["dependency-skill-id"]

Full guide at docs/custom-skills.md.

---

Development

npm run typecheck    # TypeScript strict mode
npm run lint         # ESLint
npm test             # Jest (unit + integration + E2E)
npm run build        # Compile to dist/
npm run dev -- init "My idea"   # Run in dev mode

Stack: TypeScript, Commander.js, Inquirer.js, Ora, Chalk, fs-extra, Handlebars, Zod, Jest, EventEmitter3.

---

Contributing

Contributions welcome! See CONTRIBUTING.md and ROADMAP.md.

High-impact areas:

• Custom executors — Wire up real LLM providers (OpenAI, Anthropic, Ollama, Gemini, etc.)
• New skill definitions — Add YAML for additional engineering roles
• Validation stages — Extend the pipeline with new checks
• Storage backends — Add S3, DynamoDB, PostgreSQL adapters
• UI dashboard — Web interface for workflow visibility

---

Show Your Support

---

License

MIT — Copyright (c) 2026 forgephantom. See LICENSE for full text.

---

Legal Disclaimer

AI Skill Engineer is a development tool that generates code and project artifacts using AI models. By using this software:

1. AI-generated code — The output produced by this tool is generated by large language models and may contain errors, security vulnerabilities, or non-compliant code. You are responsible for reviewing, testing, and validating all generated output before deploying it to any production environment.

2. No warranty — The software is provided "AS IS", without warranty of any kind. The generated output is not guaranteed to be correct, secure, performant, or free of defects.

3. --no-human-approval — This flag bypasses the human review gate. Use it only for development/testing. For production use, always review generated code manually and ensure proper security and compliance checks.

4. Third-party dependencies — The generated project may include suggestions for third-party packages, libraries, or services. You are responsible for reviewing their licenses and terms of service.

5. API keys — You are responsible for all API usage and costs associated with your own API keys. The maintainer does not provide or manage API keys.

6. Compliance — You are solely responsible for ensuring that any project generated with this tool complies with all applicable laws, regulations, and industry standards.