A skill that hands coding work to OpenAI Codex through a dedicated connection. It supports implementation, planning, questions, code review, and continuing earlier delegations.

Use it to delegate multi-file features, refactors, implementation plans, advisory questions, or reviews of code changes to Codex.

It removes the need to perform larger coding tasks in the current workflow and provides a structured handoff with scope and acceptance criteria.

delegate — Skill by andreilungeanu

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods

npx agentmods add skills/andreilungeanu/codex-delegate-mcp/delegate

Any agent

npx skills add andreilungeanu/codex-delegate-mcp --skill delegate

Clone the repo

git clone --depth 1 https://github.com/andreilungeanu/codex-delegate-mcp

Made for: Claude Code, Codex.

Or install codex-delegate, the plugin that ships this one along with the rest of its 1 skill, 1 command, 1 MCP server.

Per session 83 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 1,107 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5	$0.00083	$0.01107
Opus 5	$0.00042	$0.00553
Sonnet 5	$0.00017	$0.00221
Haiku 4.5	$0.00008	$0.00111

Measured 2d ago against content hash 3aed4775ca52, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

delegate scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/delegate/SKILL.md · 79 lines

How it starts

The opening of the file, as written. The whole thing — 79 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Delegate to Codex

You orchestrate; Codex implements. Use the codex-delegate-mcp MCP server — never run codex from the shell for these tasks.

For the full input and result field semantics, the mode rules, and how concurrency and timeouts behave, read reference.md in this skill directory.

When to delegate

Trivial (one-liner, rename, typo): do it yourself.
Medium (multi-file feature or refactor): one delegate call.
Large or risky: mode: "plan" first; implement after approval.
Advisory questions → mode: "ask". Code review of a diff/commit → mode: "review".

Workflow

Build the brief inline in spec:
- Goal — the outcome, precisely.
- Scope — which files/directories are in play.
- Decisions already made — quote the user's exact values verbatim.
- Done when — verifiable acceptance criteria. Point at files to read; don't paste large code blocks.
Call delegate on codex-delegate-mcp.
Review — check status before trusting result: a run that spawns and then fails returns normally rather than raising, so a caller that only catches errors reads a failure as an empty success. Then read warnings, then filesReportedByEditTools, then the git diff; run tests/lint. The field lists only what Codex's edit tool reported — files it wrote through a shell command are missing, and anything it edited outside the workspace is listed as an absolute path — so the diff is the better record, not a complete one.
- A warnings entry always means something real; empty warnings is not a clean bill of health — the bridge sees only failures Codex reports as failed or declined tool calls, not ones it narrates in result. Such a warning reports Codex's status, not a verdict; it carries its own reading of that status.
- result is the authoritative final message only on status: "completed"; every other status carries an empty result and a reason. result-unavailable means the run finished but its final message could not be read or parsed — inspect the diff and resume the thread for a concise final answer.
- If criteria fail: resume the same thread with resumeThreadId and a specific fix brief (pass the same workspace).
- If a resume returns resumed: false with a new threadId, Codex minted a fresh thread and prior context did not carry over.
- After 2 failed resumes, start a fresh thread with a rewritten brief.
Report — summarize what changed and whether acceptance criteria are met.

Read the full file on GitHub · 79 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

reference.md 6.2 KB

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

2d ago First seen · 79 lines · 83 tokens per session scan A 3aed4775ca52

Subscribe to this mod's changes

delegate is a skill published in the GitHub repository andreilungeanu/codex-delegate-mcp (1 stars, last pushed 12d ago), licensed MIT. It adds 83 tokens to every session and 1,107 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

agents ai-agent ai-agents claude-code claude-plugin codex

ork-assess

Assess a code change, design, architecture, workflow, or competing options against explicit criteria and evidence. Use when a request asks to assess, rate, compare, identify trade-offs, evaluate readiness, or decide whether an approach is good enough. Do not use for a full pull-request review or to implement a chosen…

yonatangross/orchestkit · 68 tokens

ork-brainstorm

Compare plausible implementation, architecture, product, or operational approaches before committing to one. Use when a request asks to brainstorm, think through options, choose an approach, evaluate trade-offs, or resolve significant uncertainty before implementation. Do not use for an already specified mechanical…

yonatangross/orchestkit · 58 tokens

ork-explore

Map an unfamiliar codebase, feature, architecture, data flow, or operational path with file-backed evidence. Use when a request asks how a system works, where behavior lives, what changed, which dependencies matter, or for onboarding before a change. Do not use to implement or fix the code.

yonatangross/orchestkit · 63 tokens

ork-review-pr

Review a pull request or branch for correctness, regressions, security, operational risk, and missing evidence. Use when a request asks to review a PR, review a diff, find real bugs, assess merge risk, or provide evidence-backed review findings. Do not use for implementation or style-only cleanup.

yonatangross/orchestkit · 64 tokens

ork-verify

Verify that existing work is ready to merge, release, or hand off using an explicit evidence contract. Use when a request asks to verify, validate, prove, check readiness, run the relevant tests, or distinguish a claimed result from an observed one. Do not use to write missing tests or fix failures.

yonatangross/orchestkit · 65 tokens

ork-implement

Make an approved, scoped change and prove the affected behavior. Use when a request asks to implement, build, add, or land a feature that already has an agreed approach. Do not use to explore, review, or verify existing work, or to choose between approaches.