# How Addy Osmani actually uses coding agents

Claude Code at Anthropic, ex-Chrome

**If you take one thing:** `/loop PR triage`, on Agent Skills, every day. scheduled agent loops make the first pass on the PR queue of his 100k-star skills repo.

- **Agent:** Claude Code ([source, 2026-08-14](https://addyo.substack.com/p/practical-loop-engineering))
- **Agent:** Codex ([source, 2026-06-16](https://addyo.substack.com/p/agentic-code-review))
- **Model:** Fable ([source, 2026-09-08](https://x.com/addyosmani/status/2097210828659564642))
- **Review:** Careful human review of core paths only ([source, 2026-09-28](https://addyo.substack.com/p/the-code-nobody-reads))
- **At once:** 5 at once ([source, 2026-08-14](https://addyo.substack.com/p/practical-loop-engineering))
- **Harness:** Claude Code, a git worktree per feature ([source, 2025-12-18](https://addyo.substack.com/p/my-llm-coding-workflow-going-into))
- **Rules file:** Personal CLAUDE.md, audited with /doctor ([source, 2026-08-27](https://addyo.substack.com/p/audit-your-agent-files))
- **Skills:** api-and-interface-design, browser-testing-with-devtools, ci-cd-and-automation, code-review-and-quality, code-simplification, constraint-driven-development, context-engineering, debugging-and-error-recovery, deprecation-and-migration, documentation-and-adrs, doubt-driven-development, frontend-ui-engineering, git-workflow-and-versioning, idea-refine, incremental-implementation ([source, 2026-10-03](https://github.com/addyosmani/agent-skills/tree/1401c8b8030e023baeebb31781a6653fe8e93026/skills))
- **MCP servers:** context7, chrome-devtools ([source, 2025-12-18](https://addyo.substack.com/p/my-llm-coding-workflow-going-into))

## In their words

> I can now largely rely on the output of the primitives in Claude Code and Codex.

> I point Claude Code or Codex at a batch of incoming PRs and ask for a first pass: a high-level read of what looks safe to merge, what needs more work, and what is genuinely high-risk.

> Here's a little Fable & Three.js

> For me, I use probably between five and ten agents every day. Very typically I’ll max out at about five concurrently.

> Core and sensitive paths still need an owner and a careful human review, and that’s where I spend my attention.


Archetype (our label): Orchestrator. Newest source: 2026-10-03.
Card: https://www.prehistoriq.com/stacks/addyosmani/ · Data: https://github.com/prehistoriq/stacks/blob/main/data/stacks/addyosmani.json