Compound Engineering

Why pick it

Most agent plugins give you one-off commands. Compound Engineering gives you a loop that gets smarter every cycle: brainstorm → plan → work → simplify → review → compound — and each turn feeds learnings back so the next task starts with more context, not less.

The core idea: 80% of the value is in planning and review, 20% in execution. Instead of letting an agent rush to write code (and accumulate debt), CE makes it think first, ship cleaner, and write down what it learned.

What you get

• /ce-brainstorm — interactive Q&A to nail down what to build before planning.
• /ce-plan — turn a rough idea into an implementation-ready plan.
• /ce-work — execute the plan, with host-side verification and commits.
• /ce-simplify-code — clean up what was just written before review.
• /ce-code-review — multi-agent review against the plan before merge.
• /ce-compound — capture learnings into docs/solutions/ so the next loop is sharper.

Why it over a plain skill pack

vs wshobson/agents or gstack — those are collections of independent slash commands. CE is a connected workflow: each skill reads what the previous one produced, and /ce-compound writes knowledge the next /ce-brainstorm reads. The loop compounds; a command list doesn't.

vs Ponytail — Ponytail pushes agents toward the smallest solution (YAGNI). CE is complementary: it adds the planning and review discipline around execution. Many users run both.

Works with

Claude Code, Codex (CLI + app), Cursor, Cline, Grok, Devin, GitHub Copilot, OpenCode, and more — install from each host's plugin marketplace.

Quick start

In Claude Code: /plugin marketplace add EveryInc/compound-engineering-plugin then /plugin install compound-engineering. Start with /ce-brainstorm on your next feature.

23k+ stars, 900+ contributors. The philosophy: each unit of engineering work should make the next one easier, not harder.