codex

A command that sends a task to a separate Codex agent through the Clanker system. It can run the task in read-only mode or, with the write option, in an isolated worktree.

In plain words
What is it for?
Use it to pass coding tasks to Codex with optional model and reasoning-effort settings. It relays the agent's progress and final result back to you.
Why use it?
It lets you delegate work while keeping the main checkout protected. Read-only mode is the default, so the agent cannot change files unless writing is explicitly enabled.

Command

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 commands/kckylechen1/clankerhouse/codex
Clone the repo
git clone --depth 1 https://github.com/kckylechen1/ClankerHouse
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 704 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 $0.00020 $0.00704
Opus 5 $0.00010 $0.00352
Sonnet 5 $0.00004 $0.00141
Haiku 4.5 $0.00002 $0.00070

Measured yesterday against content hash edc164114314, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

codex 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 yesterday.

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.

Origin

Copies of this mod

2 near-identical copies found in the catalogue:

  • grok — 92% identical, 20 lines differ
  • oc — 81% identical, 26 lines differ
plugin/commands/codex.md · 39 lines

What it actually says

Dispatch the task below to the Clanker: Codex by invoking the clanker:codex subagent via the Agent tool (subagent_type: "clanker:codex"). The subagent is a zero-discretion relay: it calls clanker_dispatch_start once and long-polls clanker_wait until the turn completes, its transcript showing the progress digests. Its final message (final_message + result fields) is what you relay to the user.

Raw request: $ARGUMENTS

Argument mapping (resolve these, then pass explicit parameters to the subagent):

  • --write present → read_only: false and a worktree is mandatory (the server rejects a write without one — writes must be isolated from the primary checkout). If the user did not name a branch, generate a default worktree like clanker/codex-<short-timestamp>. Absent --write (or --read-only present) → read_only: true (safe default; the Clanker cannot modify files, no worktree needed).
  • --model <model>model (passed through verbatim; codex maps it via CODEX_CONFIG).
  • --effort <effort>effort (codex reasoning effort).
  • Everything that is not a recognized flag is the natural-language prompt. Do not forward the flags themselves as prompt text.
  • --background and --wait are Claude execution flags. Do not forward them to the Clanker prompt, and do not treat them as part of the natural-language task text.

Execution mode:

  • If the raw request includes --wait, run the clanker:codex subagent in the foreground and relay its final result verbatim.
  • If the raw request includes --background, run the clanker:codex subagent as a Claude Code background task with run_in_background: true.
  • If neither flag is present, default to background. The whole point of this command is that Claude Code owns the visible Clanker task row; the MCP server only owns the ACP backend.

Background flow:

  • Launch exactly one Agent tool call with subagent_type: "clanker:codex" and run_in_background: true.
  • The task text must be: "Dispatch as this Clanker. prompt=. read_only=. worktree=. model=. effort=. Follow your relay protocol."
  • Do not call MCP dispatch tools in the main conversation.
  • Do not wait for the subagent or relay a final result in this turn.
  • After launching, tell the user: "Clanker: Codex started in the Claude Code background task list."

Foreground flow:

  • Launch exactly one Agent tool call with subagent_type: "clanker:codex" and no background flag.
  • Use the same task text as the background flow.
  • Relay the subagent's final result verbatim.
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. yesterday First seen · 39 lines · 0 tokens per session scan A edc164114314

Subscribe to this mod's changes

codex is a command published in the GitHub repository kckylechen1/ClankerHouse (0 stars, last pushed 1mo ago), licensed MIT. It adds 20 tokens to every session and 704 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.