Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add bengous/claude-code-plugins/plugin install agents-bridgeWrote 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.
[](https://agentmods.dev/skills/bengous/claude-code-plugins/codex-orchestrate)<a href="https://agentmods.dev/skills/bengous/claude-code-plugins/codex-orchestrate"><img src="https://agentmods.dev/badge/skills/bengous/claude-code-plugins/codex-orchestrate/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.
<a href="https://agentmods.dev/skills/bengous/claude-code-plugins/codex-orchestrate"><img src="https://agentmods.dev/badge/skills/bengous/claude-code-plugins/codex-orchestrate.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00076 | $0.01897 |
| Opus 5.5 | $0.00030 | $0.00759 |
| Sonnet 5.5 | $0.00015 | $0.00379 |
| Haiku 4.5 | $0.00008 | $0.00190 |
Grade A, and why
codex-orchestrate 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 14d 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.
How it starts
The opening of the file, as written. The whole thing — 142 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Orchestrating a plan in slices through Codex
Claude Code only. In native Codex, use $slice-runner.
Principle
Claude is architect, QA and committer; Codex is the executor. Claude cuts the plan into sequential slices, imposes the interfaces, starts one Codex run per slice, checks the gates itself, and commits. Codex never commits and never designs: if Claude lets Codex invent an API, N slices produce N styles.
Model routing per slice
Pick the role when writing the slice's prompt:
| Slice | Role | Why |
|---|---|---|
| Standard code (features, wiring, specified refactor) | coding |
Default. |
| Tests, mechanical fixes to pass a gate | bounded |
Fastest and cheapest, enough for bounded work |
| Code that needs judgment or rigour, dense tricky points | audit |
The highest ceiling |
"${CLAUDE_PLUGIN_ROOT}/scripts/codex-run.ts" models prints the model and
effort behind each role and every effort each model supports. Hard decisions
(architecture, API choices) stay Claude's work — a slice that contains one is a
slicing defect, not a reason to move up a tier.
These models delegate well themselves: on a heavy slice, allow the prompt to hand mechanical sub-parts to a cheaper model, rather than over-slicing on the orchestrator's side.
Context rule (non-negotiable): never read a Codex run's full event stream or
its full diff. Reading = the final message (Codex's summary) + git diff --stat
- gates. Targeted inspection (grep, partial Read) only when a gate fails or the slice has a risk named in advance in its prompt.
Pre-flight
- A plan the user validated, cut into slices: small, sequential, each with an
explicit exit gate (e.g.
validatealone for internal wiring; + e2e/visual for what touches the DOM). - Slice 0 = baseline: every gate green before the first slice. Otherwise, stop and report.
- Branch: follow the user's instruction; by default a dedicated branch when a push to main triggers something. Never push — the user pushes.
- First slice = the smallest and most self-contained one (a smoke test of the pipeline: environment, conventions, sandbox).
TaskCreateone task per slice;TaskUpdateas you go.- Insert a prerequisite slice as soon as a cross-cutting blocker shows up (e.g. test tooling missing for the new pattern) — never fold it into the current slice.
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.
- 14d ago Changed · +28 lines scan B → A 19a5866f7f06
- 1mo ago First seen · 114 lines · 76 tokens per session scan B f6a62043dec7
codex-orchestrate is a skill published in the GitHub repository bengous/claude-code-plugins (5 stars, last pushed today), licensed MIT. It adds 76 tokens to every session and 1,897 once invoked, about $0.0003 per session on Opus 5.5. 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.
Other skills, from other repositories
html-ppt-hermes-cyber-terminal
OpenDesign + BYOK: choosing and wiring your own model, hands-on — cost, quality, and the routing decision. Built as a decision-grade AI literacy deck for engineers, IT, applied-AI teams.
development
An index of programming guidance for Python, Go, Rust, TypeScript, Java, C++, and shell scripting.
tencent-docs
A Tencent Docs assistant for creating, reading, editing, organizing, importing, and exporting online documents, spreadsheets, presentations, diagrams, and forms.
wechat-search
Search WeChat Official Account articles using OpenClaw's web search, Tavily API, and web fetch capabilities with compliance-focused design.
debugging-executions
Debug failed or wrong-output workflow executions using executions tools. Load when the user reports execution failures, unexpected node output, empty parameter values after a successful run, or a node showing a red or failed expression error. Also load when a live run you started for the user fails with an error from…
flow-next-tracker-sync
Project a flow-next spec to a tracker issue (Linear, GitHub, GitLab, Jira) and reconcile two-way. Use when asked to sync to a tracker. NOT plan-sync.