Swarm

What it does

When your agent uses it

• Tasks that involve Subagents
• Tasks that involve Git worktrees

Example prompts

• “/swarm”

Requirements

• Pre-approved tools (allowed-tools): Bash, Read, Write, Edit, Glob, Grep, Agent, TeamCreate, TeamDelete, SendMessage, TaskCreate, TaskList, TaskUpdate, TaskGet, TaskOutput, TaskStop

Workflow steps

5 steps, taken from the first numbered list in SKILL.md.

1. 1Collect all agent outputs into a uniform format (JSON or Markdown sections)
2. 2De-duplicate identical findings across agents
3. 3Synthesize unique insights into one report
4. 4For Consensus mode: assess each proposal against shared evidence and acceptance criteria; investigate conflicting findings and record…
5. 5Generate final swarm report

What it can do on your machine

Read from SKILL.md and the folder at commit d64db2b. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves these tools, so the agent can use them without asking each time:

  • Bash
  • Read
  • Write
  • Edit
  • Glob
  • Grep
  • Agent
  • TeamCreate
  • TeamDelete
  • SendMessage

  …and 6 more on the same allowed-tools line.

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  Shell commands in SKILL.md call:

  • git

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  No URLs in SKILL.md. Its commands use git, which can reach the network depending on how they are called.

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Swarm loads about 1.9k tokens when it runs. Until then it costs about 34 tokens; SKILL.md has 711 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~34

When it runs · the whole SKILL.md, loaded when a task matches

~1.9k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

• NotePre-approves every shell command (allowed-tools: Bash)SKILL.md

  allowed-tools: Bash, Read, Write, Edit, Glob, Grep, Agent, TeamCreate, TeamDelete, SendMessage, TaskCreate, TaskLis

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from softspark/ai-toolkit at commit d64db2b, republished under its Apache-2.0 licence (© softspark). 711 words, ~1,878 tokens.

name

swarm

description

Runs tasks via Map-Reduce, Consensus, or Relay swarms. Triggers: swarm, map-reduce, consensus swarm, relay swarm, parallel agents.

allowed-tools

Bash, Read, Write, Edit, Glob, Grep, Agent, TeamCreate, TeamDelete, SendMessage, TaskCreate, TaskList, TaskUpdate, TaskGet, TaskOutput, TaskStop

user-invocable

true

effort

high

argument-hint

[map-reduce|consensus|relay] [--with-kb] [--worktree] [task]

context

fork

agent

orchestrator

model

opus

<!-- CLAUDE_CODE_ONLY_START -->

<!-- CLAUDE_CODE_ONLY_END -->

# Single response with N Agent tool calls:
Agent(subagent_type="...", prompt="sub-task 1 — own files: path/a/")
Agent(subagent_type="...", prompt="sub-task 2 — own files: path/b/")
Agent(subagent_type="...", prompt="sub-task N — own files: path/n/")

Agent(subagent_type="backend-specialist",      prompt="[problem] — approach from data layer angle. Output: solution + confidence 0.0–1.0")
Agent(subagent_type="tech-lead",               prompt="[problem] — approach from architecture angle. Output: solution + confidence 0.0–1.0")
Agent(subagent_type="performance-optimizer",   prompt="[problem] — approach from performance angle. Output: solution + confidence 0.0–1.0")

# Round 1
Agent(subagent_type="tech-lead", prompt="Design the API spec. Output to docs/api-spec.md")
# Wait for completion

# Round 2
Agent(subagent_type="backend-specialist", prompt="Implement based on docs/api-spec.md. Own files: src/")
# Wait for completion

# Round 3
Agent(subagent_type="test-engineer", prompt="Write tests for src/. Own files: tests/")

Agent(
  subagent_type="<agent-name>",
  description="<3-5 word summary>",
  prompt="<full task description including: original request, specific sub-task, owned files, success criteria>"
)

[KB CONTEXT — from rag-mcp smart_query, ground all decisions in these]
- {kb_id}: {title}
  {content excerpt, ≤300 chars}
- ...

[YOUR SUB-TASK]
{specific sub-task, owned files, success criteria}

[RULES]
- Cite KB entries as [PATH: kb_id] when you rely on them.
- If KB is silent on a decision, state that explicitly — do NOT invent.
- After producing your output, call mcp__rag-mcp__verify_answer with your answer + the cited kb_ids; include the verdict in your final report.

Agent(
  subagent_type="...",
  description="...",
  prompt="...",
  isolation="worktree"
)

© softspark, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in app/skills/swarm of softspark/ai-toolkit.

Open the folder on GitHubat commit d64db2b