What is codex-claude-worker?

A workflow for turning a natural-language coding request into a limited task for Claude Code, while Codex or ChatGPT supervises the work.

Use it to inspect a repository, prepare bounded worker prompts, run approved tasks, read worker summaries, and start a focused fix when validation finds a defect.

It defines the worker's scope, forbidden actions, expected files, and checks, then requires the supervisor to review the result independently.

codex-claude-worker — Skill for Codex by zhuzai-2007

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.

Any agent

npx skills add zhuzai-2007/claude-code-mcp-harness --skill codex-claude-worker

Clone the repo

git clone --depth 1 https://github.com/zhuzai-2007/claude-code-mcp-harness

Made for: Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/skills/zhuzai-2007/claude-code-mcp-harness/codex-claude-worker/github.svg)](https://agentmods.dev/skills/zhuzai-2007/claude-code-mcp-harness/codex-claude-worker)

Your own site

<a href="https://agentmods.dev/skills/zhuzai-2007/claude-code-mcp-harness/codex-claude-worker"><img src="https://agentmods.dev/badge/skills/zhuzai-2007/claude-code-mcp-harness/codex-claude-worker/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/skills/zhuzai-2007/claude-code-mcp-harness/codex-claude-worker"><img src="https://agentmods.dev/badge/skills/zhuzai-2007/claude-code-mcp-harness/codex-claude-worker.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 86 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 986 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

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 5.1	$0.00086	$0.00986
Opus 5	$0.00043	$0.00493
Sonnet 5	$0.00017	$0.00197
Haiku 4.5	$0.00009	$0.00099

Measured 10d ago against content hash faaf7c99cc42, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

codex-claude-worker 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 10d ago.

The scan reads SKILL.md. This mod also ships 1 executable file (scripts/verify-harness.ps1), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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/codex-claude-worker/SKILL.md · 76 lines

How it starts

The opening of the file, as written. The whole thing — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Codex Claude Worker

Use this skill when Codex or ChatGPT is the main reasoning agent and Claude Code is a bounded local worker. The common input is a normal user prompt, not a complete specification. Treat the harness as a policy and workflow boundary, not an OS sandbox.

Operating Model

Translate the user's natural-language request into a concrete worker task: goal, target path, deliverables, forbidden actions, acceptance checks, and budget.
Inspect the repo and choose the smallest safe worker boundary.
Use plan for read-only orientation when task scope or write boundary is uncertain.
Use run only with explicit approval metadata and a bounded prompt.
Read summary.ps1, worker-result.normalized.json, and the ledger after every run.
Independently validate worker output before reporting success, including artifact_status and supervisor_notes when the worker failed or hit budget.
Start a separate fix run when validation finds a concrete defect.

Prefer direct PowerShell harness use first. Treat the MCP bridge as optional transport, not the only interface.

Auto-Run Rules

Auto-run is acceptable only for small tasks that create or modify files under one new directory such as workspace/<task-id>/.

Require a second user confirmation before run mode when the task touches existing source, .agents/, mcp-server/, scripts/, config files, package files, lockfiles, dependencies, network, deletion, git operations, broad scans, long tasks, or ambiguous write boundaries.

Always include these boundaries in worker prompts:

Only create or modify files under <bounded-path>.
Do not modify .agents/, mcp-server/, scripts/, .git/, config files, package files, or lockfiles.
Do not access the network.
Do not install dependencies.
Do not delete files outside <bounded-path>.
Do not run git commands.
Return concise JSON with summary, files_read, changes_made, commands_run, tests_or_checks, risks, blocked_on.

Prompt Translation

Read the full file on GitHub · 76 lines

Files

What ships with it

5 files 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.

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.

10d ago First seen · 76 lines · 86 tokens per session scan A faaf7c99cc42

Subscribe to this mod's changes

codex-claude-worker is a skill published in the GitHub repository zhuzai-2007/claude-code-mcp-harness (8 stars, last pushed 1mo ago), licensed MIT. It adds 86 tokens to every session and 986 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.

Building agents Claude Code LLM automation coding-agent developer-tools mcp

source-research

Research open-source libraries with evidence-backed answers and GitHub permalinks. Use when the user asks about library internals, needs implementation details with source code references, wants to understand why something was changed, or needs authoritative answers backed by actual code. Excels at navigating large…

aretw0/agents-lab · 71 tokens

provider-model-discovery

Descobre e seleciona modelos de providers LLM de forma report-only: inventário read-only de modelos, docs oficiais, quota/billing/rate gates e shortlist para canary protegido. Use antes de adicionar providers, escolher modelos ou migrar monitores/roteamento.

aretw0/agents-lab · 60 tokens

codex-claude-worker