The Swarm — Data skill for Claude Code
Claude Code plugin for multi-pattern multi-agent orchestration, using CC's built in Agent Swarm as the engine — dispatches and coordinates specialist sub-agents in fan-out, pipeline, map-reduce, specu.
How to install The Swarm
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open ben-alkov/the_swarm and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What The Swarm does
Claude Code plugin for multi-pattern multi-agent orchestration, using CC's built in Agent Swarm as the engine — dispatches and coordinates specialist sub-agents in fan-out, pipeline, map-reduce, speculative, and swarm configurations.
Alternatives in Data
- Pipeline Feature — Add a feature to the FFmpeg transcode pipeline following Tunarr's pipeline builder pattern 2.5k ★
- Debug Plus — Enhanced debug with swarm integration and prevention pipeline 508 ★
- Graph Engineering — Graph engineering for AI agents: the 9-stage knowledge-graph pipeline (translated from SEU's graduate course) 452 ★
README
the_swarm
A Claude Code plugin which orchestrates multiple agents in parallel. Define specialist roles, pick a pattern, and let the swarm work concurrently — review code from three angles, race competing implementations, or chain stages with dependency gates.
Requirements
- [Claude Code][] v2.1.49+
bash,jq
Installation
claude plugin marketplace add ben-alkov/the_swarm
claude plugin install the-swarm
Quick Start
> /swarm review this PR with the pr-review preset
Claude reads the preset, spawns three specialists (security, performance, quality), collects their findings, and synthesizes a report — all in one command.
Presets, specialists, and even orchestration patterns (see below) are fully customizable, extensible, and composable (although the orchestration patterns generically cover a very large space).
Tips
- Jesse Vincent's [Superpowers][] works unreasonably well alongside
the_swarm. Prompt as you normally would, e.g.
/using-superpowers Review PR #whatever. Use the swarm. - Want reviewer team members that better fit your repo/stack/language? Tweak
the
promptfields inconfig/swarm-roles.yaml - Not working on source code? Create your own roles!
Orchestration Patterns
Fan-Out
Parallel specialists analyze the same target independently. Each specialist has a distinct focus; the lead merges their findings into one report.
> /swarm review src/auth/ with pr-review preset
**Built-in presets:** `pr-review`, `full-review`, `security-audit`
Pipeline
Stages run sequentially; each stage blocks the next. An implementer writes code in a worktree; reviewers evaluate the result only after implementation completes.
> /swarm implement and review with implement-then-review preset
Task Graph
An arbitrary DAG of stages with `depends_on` edges supports fan-in (multiple predecessors gate a single successor) and fan-out (one predecessor unblocks several successors).
``
Related Skills
Skill To Workflow
Compile any Claude Agent Skill into a multi-agent Workflow — a meta-workflow that recovers the parallelism hid
Fanout Loop
Execute ONE iteration of an effort's fan-out loop — pick-up, derive the run, dispatch a tier-laddered pipeline
FACT
FACT – Fast Augmented Context Tools: FACT is a lean retrieval pattern that skips vector search. We cache every
Synagon
A multi-agent AI orchestrator that coordinates local coding CLI agents (Claude Code, OpenCode, Antigravity, Co
Crucible
Multi-agent adversarial research swarm. LLM agents compete to optimize ML artifacts -- and both independently
Agentic Document Extraction
Agentic Document Extraction (ADE) pipeline built from scratch on Claude. Upload a PDF, describe in natural lan
Related Agents
Kimi Swarm Write
Use this agent ONLY when the user has explicitly asked Kimi to make EDITS across MANY disjoint targets IN PARA
Engine Grounding Scout
Use when you are about to run the memory-recall sweep or a knowledge-graph impact traversal and the sweep will
Deep Research Sonnet
Bulk-fan-out worker for multi-axis breadth-first research. Sonnet 5 tier (NOT frontier) — the canonical worker