Skills for delegating coding work to a separate CLI agent and landing it yourself. Your agent (the orchestrator) writes a self-contained brief, hands it to an implementer CLI, then reviews the diff and commits — staying the reviewer the whole way.
The first skill, codex-delegate, drives the OpenAI Codex CLI.
Install
Browse first:
npx skills add amElnagdy/delegate-skills --list
Install the package, or just one skill:
npx skills add amElnagdy/delegate-skills
npx skills add amElnagdy/delegate-skills --skill codex-delegate
Install for a specific agent, or globally:
npx skills add amElnagdy/delegate-skills --skill codex-delegate --agent claude-code
npx skills add amElnagdy/delegate-skills --global
Works with any orchestrating agent the Skills CLI supports.
What it does
The loop, for codex-delegate:
- Write a brief — a self-contained task spec; Codex sees only what you send.
- Dispatch it with the bundled
relay.mjs(a thincodex execwrapper). - Wait for completion — the helper writes a structured
result.json. - Review the diff — re-run the project's gates yourself; pair with guard skills.
- Land it — you commit, because the implementer's sandbox can't reliably write
.git.
Use $codex-delegate to have Codex implement the refactor in services/billing/, then review and commit it.
Use $codex-delegate to run this queue of migration tasks through Codex while I review each one.
How this differs from the OpenAI Codex plugin
The official openai-codex Claude Code plugin is excellent and
complementary — this skill builds on the same codex CLI, it doesn't replace the plugin. They
point in different directions:
- The plugin's
codex:codex-rescueagent is a forwarder: it hands one task to Codex and returns the output. It deliberately does not poll, review, or commit. - The plugin's review command and stop-review gate run the inverse direction: Codex reviews your work.
codex-delegateis the orchestration loop in the other direction: you drive Codex to implement across one task or a queue, and you review and land each result. That loop — brief → dispatch → poll → review → commit, with the orchestrator owning the commit — is what the plugin leaves to you, and what this skill encodes.
If you have the plugin installed, its companion CLI is an optional alternative dispatch backend; the
bundled relay.mjs is the default because it needs nothing but the codex binary.
The skills
codex-delegate
Drive the OpenAI Codex CLI as a background implementer: write the brief, dispatch via relay.mjs,
review the diff, commit it yourself. Ships four references (writing the brief, dispatch/poll, review/
land, multi-task queues) loaded only when needed, and one small helper script.
You'll feel it when: a bounded task — a migration, a mechanical refactor, a removal sweep — gets handed to Codex, comes back as a clean diff with a structured report, and you commit it after re-running the gates yourself instead of typing it all by hand.
gemini-delegate
Planned. A delegate skill for the Gemini CLI, if and when it gains a comparable non-interactive mode. Reserved so the umbrella can grow without a rename.
Requirements
- The
codexCLI installed and authenticated (codex login). - Node 18+ and
git. - An orchestrating agent that can run shell commands and read files.
- Shell examples assume bash/zsh (macOS/Linux, or Git Bash/WSL on Windows).
Trust and validation
This package is intentionally inspectable:
- All skill content is Markdown, plus exactly one executable:
skills/codex-delegate/scripts/relay.mjs. relay.mjsitself makes no network calls, reads or writes no credentials, sends no telemetry, and has no dependencies (Node built-ins only). It shells out only tocodexandgit. Thecodexprocess it launches authenticates exactly as you do at the terminal. Read the script before you run it.- It never commits — committing is always the orchestrator's job, after review.
Verification status: the relay's codex integration (flags, exit codes, result.json) is verified;
the end-to-end loop is designed for and run on Claude Code, but not yet formally verified across a full
delegate → review → commit cycle. Other orchestrators (OpenCode, Cursor, …) are designed-for but
unproven. This line gets upgraded to "verified end-to-end" with evidence, not assumption.
Repository shape
skills/
└── codex-delegate/
├── SKILL.md
├── scripts/relay.mjs
└── references/
├── writing-the-brief.md
├── dispatch-and-poll.md
├── review-and-land.md
└── multi-task-queues.md
The SKILL.md stays small so it loads cheaply; the references load only when the task needs them.
Comments