# EveryInc/compound-engineering-plugin

Source: [GitHub](https://github.com/EveryInc/compound-engineering-plugin)  
Feed7 permalink: https://feed7.dev/p/compound-engineering-plugin-06exf60  
Published: Unknown  
Trust: Needs Review (needs_review)

## Why Included

Compound Engineering packages a six-stage agent workflow into 32 skills for Codex, Claude Code, Cursor, and others. Its core idea is to persist plans, reviews, and solved-problem notes as future context.

## Source Summary

Every's Compound Engineering plugin ships **32 skills** and no standalone agents. Its **six-step loop** moves from requirements and planning through implementation, simplification, review, and captured learning, with artifacts stored in repository folders by default.

## Practical Implication

Builders can use the pattern without adopting every command: make requirements and plans durable, verify delegated work in the host agent, keep review report-only until fixes are approved, and record reusable solutions so later agents avoid rediscovery.

## Agent-Ready Context

Every's Compound Engineering plugin ships **32 skills** and no standalone agents. Its **six-step loop** moves from requirements and planning through implementation, simplification, review, and captured learning, with artifacts stored in repository folders by default.

Builders can use the pattern without adopting every command: make requirements and plans durable, verify delegated work in the host agent, keep review report-only until fixes are approved, and record reusable solutions so later agents avoid rediscovery.

Codex installation requires adding a custom marketplace, installing the plugin, and restarting. Existing installs need a marketplace refresh, while older Codex conversions may leave an obsolete tool-map block that must be removed only by its exact sentinels.

## Connected Context

Feed7 judgment across 353 accumulated Signals:

This turns lifecycle-skill patterns into a concrete six-step repository workflow: durable requirements, host-verified implementation, approval-gated fixes, and captured learning. Against broader skill catalogs, it favors an opinionated loop that can be adopted selectively, while its installation caveats reinforce that plugin maintenance and governance remain part of harness engineering.

- [addyosmani/agent-skills](https://feed7.dev/p/agent-skills-0d3lx1j) — Both encode the coding lifecycle as reusable skills, but this plugin adds a specific six-step loop, repository artifacts, and an explicit simplification stage.
- [Effective harnesses for long-running agents](https://feed7.dev/p/effective-harnesses-for-long-running-agents-0xzfs05) — Its durable requirements, plans, and captured learning reinforce Anthropic’s use of persistent files to carry state across agent sessions.
- [affaan-m/ECC](https://feed7.dev/p/ecc-0438yju) — This provides a narrower, workflow-centered alternative to ECC’s broad cross-agent catalog, supporting selective adoption while reducing the surface that must be verified.
- [We Vetted 2000 AI Skills Before They Reached Developers — Lucas Palma, Nubank](https://feed7.dev/p/we-vetted-2000-ai-skills-before-they-reached-developers-lucas-palma-nuba-0k4ehkz) — The installation and legacy-cleanup details make Nubank’s supply-chain framing operationally relevant: imported skills and plugin changes require controlled review and maintenance.

## Context Map

- Layer: agent
- Domains: coding
- Topics: skills, harness-engineering, agent-memory

## Uncertainty

- Codex installation requires adding a custom marketplace, installing the plugin, and restarting. Existing installs need a marketplace refresh, while older Codex conversions may leave an obsolete tool-map block that must be removed only by its exact sentinels.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.