Skip to content

README.en.md

Latest commit

 

History

History
115 lines (82 loc) · 3.21 KB

README.en.md

File metadata and controls

115 lines (82 loc) · 3.21 KB
Raw

Codex Project Command Center

中文

A Codex Skill for setting up a project workbench quickly. It separates long-running collaboration into two roles:

  • a read-only command center selects, creates, tracks, and verifies work;
  • each work task handles one feature through implementation and validation.

Project rules and durable status live in the repository. Task-level discussion stays in Codex, so work can resume in another task or on another machine without relying on chat memory.

When to use it

  • a project needs several Codex tasks over time;
  • multiple repositories, machines, or environment boundaries are involved;
  • the team wants explicit feature status, Git scope, and handoffs.

For a small one-off change, a separate command center is usually unnecessary.

Install

macOS / Linux:

git clone \
  https://github.com/Entropy-R/codex-project-command-center.git \
  "${CODEX_HOME:-$HOME/.codex}/skills/codex-project-command-center"

PowerShell:

git clone `
  https://github.com/Entropy-R/codex-project-command-center.git `
  "$HOME\.codex\skills\codex-project-command-center"

Open a new Codex task so the Skill can be discovered.

Quick start

Send this from your project:

Use $codex-project-command-center to set up a workbench for this project.
Inspect existing rules, docs, repositories, and command centers first. Ask only about
important choices that are still unresolved.

The Skill reuses existing conventions. When a choice is needed, it may ask about:

  • the repository that owns shared state;
  • the documentation location;
  • whether completed work should be committed or pushed;
  • cross-machine handoffs;
  • inherited or fixed model settings for new tasks;
  • automatic or manual command-center creation.

After confirmation, it fills only the missing documentation and creates a project-bound command center. The default documentation layout is:

AGENTS.md
docs/command-center/
├── README.md
├── features.md
└── decisions.md

Equivalent existing files are reused instead of duplicated.

Daily workflow

Typical command-center requests are:

Start the next feature.

Check the current feature.

Verify and close the current feature.

The command center remains read-only. A work task continues through analysis, implementation, tests, documentation, and the agreed Git closure. It reports back only for material choices, blockers, external authorization, handoffs, cancellation, or final completion. It never starts another feature automatically.

Repository files

  • SKILL.md: entry point and core behavior;
  • references/bootstrap.md: first-time setup;
  • references/workflow-rules.md: command-center and work-task boundaries;
  • references/templates.md: project docs, task brief, and sync templates.

Design boundaries

  • inspect before creating;
  • project rules override Skill defaults;
  • user preferences remain choices, not universal rules;
  • keep each durable fact in one authoritative place;
  • command centers do not edit code, and work tasks do not expand scope on their own;
  • commits, pushes, deployments, and other external writes follow project rules and user authorization.

License

MIT