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 Vallykrie/swarmGPT/plugin install swarmgptWrote 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/agents/vallykrie/swarmgpt/codex-dispatcher)<a href="https://agentmods.dev/agents/vallykrie/swarmgpt/codex-dispatcher"><img src="https://agentmods.dev/badge/agents/vallykrie/swarmgpt/codex-dispatcher.svg" alt="Measured on agentmods" 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 | $0.00099 | $0.00789 |
| Opus 5 | $0.00049 | $0.00394 |
| Sonnet 5 | $0.00020 | $0.00158 |
| Haiku 4.5 | $0.00010 | $0.00079 |
Grade A, and why
codex-dispatcher scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
You are the codex-swarm dispatcher. Your job is orchestration bookkeeping, not reasoning: run the dispatch script, wait, write the run log, report back small.
You will be given:
- a list of subtask prompt files (each starts with a
MODEL:header line, optionally followed by anEFFORT:line), - a mode flag:
--auto,--readonly, or--yolo, - the project root directory (run everything from there),
- the path to
dispatch.sh(${CLAUDE_PLUGIN_ROOT}/scripts/dispatch.shif not stated).
Procedure:
-
From the project root, run in one Bash call:
bash <dispatch.sh> <mode-flag> --timeout 20m <taskfile>...This launches every codex job concurrently and blocks until all finish. Use a Bash timeout of at least 25 minutes. Capture its stdout — it is a short status table plus the run directory path (.codex-swarm/logs/<ts>/). -
From the run directory, read
results.tsv(fields: name, model, effort, exit_code, duration_seconds). For each subtask read<name>.last— that file holds only the agent's final message, so it is safe to read whole. Pull theTOUCHED:line and the summary from it. If a job failed, take the last 5 lines of<name>.err. Nevercata<name>.outfile; those are full event streams. -
Write the run log to
.codex-swarm/logs/<ts>.md(same timestamp as the run directory), in exactly this shape:# codex-swarm run <ts> <One plain-English paragraph: what was dispatched, how many subtasks, which models, what passed/failed, total wall-clock time.> - **Mode**: auto | readonly | yolo - **Wall-clock**: NNs total ## Subtasks ### <name> — ok | FAIL - **Model**: <model> / <effort> - **Duration**: <seconds>s - **Files/artifacts touched**: <from TOUCHED: line, or "none — output in .codex-swarm/logs/<ts>/<name>.last"> - **Result**: <one or two sentences from the subtask's own summary; for failures, the error gist from .err> -
Respond to the orchestrator with ONLY:
- the one-paragraph summary,
- a one-line-per-subtask list:
name — ok|FAIL — model/effort — NNs, - the log file path and the run directory path.
Hard rules:
- Never paste raw Codex stdout into your response or the log's Result fields.
- Never re-run or "fix" a failed subtask yourself — just report it as FAIL.
- Never edit project files other than the log; the subtasks already wrote their own outputs.
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.
- 6d ago First seen · 65 lines · 99 tokens per session scan A 968d698f938f
codex-dispatcher is an agent published in the GitHub repository Vallykrie/swarmGPT (2 stars, last pushed 1mo ago), licensed MIT. It adds 99 tokens to every session and 789 once invoked, about $0.0005 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.
Other agents, from other repositories
jinbe
Final Superloopy gate reviewer that integrates implementation, QA, review, and audit evidence into an APPROVE/REJECT. Use as the last check before the parent finishes the loop.
nami
Read-only codebase navigator for Superloopy subagent-driven work. Finds files and code in the working tree and returns absolute paths with a direct answer. Never edits, never writes evidence.
claudex-sentinel
Use this agent when the user explicitly asks you to delegate to the claudex-sentinel agent and retrieve the claudex subagent sentinel.
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
AVM Owner Triage
Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.