不是写更长的 Prompt，而是设计更可靠的工作流。
Agentic Engineering Lab is a practical lab for exploring agentic software engineering with Claude Code, Codex, and other AI coding agents.
This repository collects workflows, plugins, skills, hooks, rules, case studies, and design notes for building reliable long-running AI-assisted development processes.
AI coding agents are powerful, but long-running software development needs more than prompt engineering.
This repository explores:
- Knowledge-first skills
- Generator-Evaluator design loops
- Checkpoint-based context handoff
- Spec-driven implementation
- Incremental deploy verification
- Hooks as hard guardrails
- Long-running change workflows
docs/
articles/
claude-code-agentic-coding-workflow.md
concepts/
spec-driven-incremental-deploy.md
theory-of-constraints-for-agentic-coding.md
claude-code/
plugins/
product-design-plugin/
The first published workflow is product-design-plugin, a Claude Code plugin for product design and implementation:
/product-design
-> requirements discovery
-> domain modeling
/architecture-design
-> architecture decisions
-> technical specification
-> design review
/implement
-> environment declaration
-> infrastructure setup
-> RED -> GREEN -> REVIEW -> DEPLOY -> COMMIT
/change
-> impact analysis
-> incremental design
-> implementation
-> regression verification
Copy the plugin into a Claude Code project:
cp -R claude-code/plugins/product-design-plugin/* /path/to/your-project/.claude/Or install only the implementation workflow:
cd /path/to/your-project
bash /path/to/agentic-engineering-lab/claude-code/plugins/product-design-plugin/scripts/setup-implement.shExperimental. Everything here is designed for learning, adaptation, and iteration.
Some assets are early workflow drafts and may require adjustment before direct production use.
MIT