codex-swarm

codex-swarm is a command for Claude Code, Codex from Vallykrie/swarmGPT. It costs 20 tokens per session (390 once invoked), scanned A, original, MIT.

A command that splits a task across several parallel Codex sessions, automatically breaking it into parts and choosing models for them.

In plain words
What is it for?
Use it to run a swarm in default, automatic, read-only, or explicitly unrestricted mode. The remaining text describes the task to distribute.
Why use it?
It helps handle tasks that benefit from separate parallel investigations or implementations, reducing the need to manage each session manually.

Command for Claude CodeCodex

Written for Claude Code and Codex: argument-hint in frontmatter, but also runs codex exec. Also seen: mentions subagents; mentions Codex.

Runs only inside its plugin — its command needs a path that Claude Code sets for a plugin’s own hooks and for nothing else. Install the plugin, not this.

Part of the swarmgpt plugin — 2 skills, 2 commands, 1 agent shipped together

Install

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.

Claude Code
/plugin marketplace add Vallykrie/swarmGPT
Claude Code
/plugin install swarmgpt

Made for: Claude Code, Codex.

Or install swarmgpt, the plugin that ships this one along with the rest of its 2 skills, 2 commands, 1 agent.

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.

agentmods badge for codex-swarm

README.md
[![agentmods](https://agentmods.dev/badge/commands/vallykrie/swarmgpt/codex-swarm.svg)](https://agentmods.dev/commands/vallykrie/swarmgpt/codex-swarm)
Your own site
<a href="https://agentmods.dev/commands/vallykrie/swarmgpt/codex-swarm"><img src="https://agentmods.dev/badge/commands/vallykrie/swarmgpt/codex-swarm.svg" alt="Measured on agentmods" height="20"></a>
Per session 20 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 390 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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00020 $0.00390
Opus 5 $0.00010 $0.00195
Sonnet 5 $0.00004 $0.00078
Haiku 4.5 $0.00002 $0.00039

Measured 5d ago against content hash e67ef00915f8, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

codex-swarm 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 5d 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.

commands/codex-swarm.md · 32 lines

What it actually says

Run the codex-swarm skill (skills/codex-swarm/SKILL.md in this plugin — read it now and follow it exactly).

Arguments given: $ARGUMENTS

Interpret them as follows:

  • If the first word is default, auto, readonly, or yolo, that is the sandbox mode; everything after it is the task description.
  • Otherwise the mode is default and the whole argument string is the task.
  • Mode meanings (details in the skill's Step 1):
    • default — mirror your own current permission mode: auto-accept host → dispatch with --auto; review-mode host → dispatch with --readonly.
    • auto--sandbox workspace-write: Codex may write inside the project.
    • readonly--sandbox read-only: analysis and research only.
    • yolo--dangerously-bypass-approvals-and-sandbox. Only on explicit request, in an already-isolated environment.

If no task description remains, treat the mode as set for this conversation, confirm it in one line, and apply it when the user gives the next swarmable task. Do not ask which model to use for any subtask — routing is your job.

Decompose the task, route models, dispatch all subtasks in parallel via the codex-dispatcher subagent (it calls ${CLAUDE_PLUGIN_ROOT}/scripts/dispatch.sh), review the results, ensure the run log exists at .codex-swarm/logs/<ISO-timestamp>.md, then integrate, verify, and report.

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.

  1. 5d ago First seen · 32 lines · 0 tokens per session scan A e67ef00915f8

Subscribe to this mod's changes

codex-swarm is a command published in the GitHub repository Vallykrie/swarmGPT (2 stars, last pushed 1mo ago), licensed MIT. It adds 20 tokens to every session and 390 once invoked, about $0.0001 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.