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.
npx skills add adriancmurray/swarm --skill using-swarmgit clone --depth 1 https://github.com/adriancmurray/swarmWrote 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/adriancmurray/swarm/using-swarm)<a href="https://agentmods.dev/skills/adriancmurray/swarm/using-swarm"><img src="https://agentmods.dev/badge/skills/adriancmurray/swarm/using-swarm/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/adriancmurray/swarm/using-swarm"><img src="https://agentmods.dev/badge/skills/adriancmurray/swarm/using-swarm.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 | $0.00048 | $0.01610 |
| Opus 5.5 | $0.00019 | $0.00644 |
| Sonnet 5.5 | $0.00010 | $0.00322 |
| Haiku 4.5 | $0.00005 | $0.00161 |
Grade A, and why
using-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 2mo 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 — 171 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Using swarm
swarm is a command-line orchestrator. You (the host agent — Claude Code,
Codex, Gemini CLI, or any other) invoke it to coordinate other agents:
fan a task out to parallel workers, run a structured multi-role discussion,
or hand a manager a pile of context to plan and synthesize. The engine routes
each role to a configured backend (a CLI agent, an HTTP API, or an in-process
harness) and returns the combined result.
Cardinal rule
This skill drives the orchestrator. It is never injected into a worker.
There are two unrelated skill layers (see the project README, "Two layers of
skills"). One layer is SKILL.md files that a native backend loads into a
single worker's system prompt. This is the other layer: a playbook for the
host agent calling the swarm binary. They never touch. Nothing you read here
changes how any worker behaves — it only tells you how to launch the swarm.
When to reach for swarm
- Multi-perspective review — you want a security lens, a simplification
lens, and a tests lens on the same diff. →
audit - Parallel fan-out — a task splits cleanly into independent slices and you
want them worked at once, then merged. →
fanout - Structured debate — two or more roles should argue a design across
rounds, not just answer in isolation. →
discuss - Manager synthesis — you have a large context and want one capable agent
to plan or reconcile it. →
metadirector - Design review — a product/UI change needs visual, motion, interaction,
and accessibility eyes. →
design
For a single routed task with no coordination, just use run (or omit the
verb — a bare prompt is treated as run).
Core verbs
Flags below are verified against the binary. A bare prompt defaults to run.
--timeout is in seconds (default 300). --cwd sets the working directory.
fanout (alias: swarm) — parallel workers + synthesis
swarm fanout [--manager AGENT[:MODEL]] [--worker ROLE=AGENT[:MODEL]]... \
[--parent ID] [--slice ID] [--cwd CWD] [--timeout SECONDS] PROMPT
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.
- 2mo ago First seen · 171 lines · 48 tokens per session scan A 36903ab7e82b
using-swarm is a skill published in the GitHub repository adriancmurray/swarm (12 stars, last pushed 4mo ago), licensed Apache-2.0. It adds 48 tokens to every session and 1,610 once invoked, about $0.0002 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 2026-08-30.
using-swarm compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| using-swarmadriancmurray/swarm | Skill | 12 | 4mo ago | Apache-2.0 | A | 48 | Claude Code, Codex |
| imagegenopenai/codex | Skill | 133k | yesterday | Apache-2.0 | A | 113 | Codex |
| playwright-climicrosoft/playwright | Skill | 99k | yesterday | Apache-2.0 | A | 19 | Claude Code |
| local-ai-agentsmicrosoft/ai-agents-for-beginners | Skill | 79k | 4d ago | MIT | A | 200 | Claude Code, Codex |
| systematic-debuggingobra/superpowers | Skill | 313k | 3d ago | MIT | A | 21 | Claude Code |
| next-cache-components-adoptionvercel/next.js | Skill | 143k | yesterday | MIT | A | 95 | Claude Code |
| next-partial-prefetching-adoptionvercel/next.js | Skill | 143k | yesterday | MIT | A | 103 | Claude Code |