// Software stack
AI coding workflow
Daily loop of reading, editing, PRs and review with coding agents.
4 pieces4 setup stepsscenario: Code with AI: choose tools, keep quality
What goes in it
- 1
Cursor Coding agents & AI IDEs Closed / commercial
IDE / agent
AI code editor and coding agent with multi-model support, background agents and review.
Or: Claude Code · Windsurf · Google Antigravity
- 2 ★ 109k
addyosmani/agent-skills Skills & rules Open source
Skill packs
Production-grade engineering skills for AI coding agents.
Or: andrej-karpathy-skills · ponytail · impeccable
- 3 ★ 48k
open-code-review Coding agents & AI IDEs Open source
Code review
Alibaba's hybrid code-review tool combining rules and an LLM agent.
- 4 ★ 32k
code-review-graph Code intelligence Open source
Code context
Local-first code intelligence graph for MCP and CLI.
Or: GitNexus · codebase-memory-mcp
How to set it up
- 01Pick one main harness (Cursor or Claude Code)
- 02Install 3–5 core skills
- 03Add a code-graph MCP server
- 04Require an agent review on every PR
Similar stacks: what’s different
- Coding agent workbenchThat one is centred on terminal agents (Claude Code + Codex)