swarm

An agent skill by cursor, from cursor/plugins. Tags: analysis, automation, multi-agent, research, workflow.

What it does

Fan out N parallel workers, drain them, and return one report. Use for /swarm, 'swarm this', or parallel coverage, races, gauntlets, and exploration.

Install

With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:

npx skills add cursor/plugins --skill swarm

Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):

git clone --depth 1 https://github.com/cursor/plugins
cp -r plugins/pstack/skills/swarm ~/.claude/skills/swarm

Safety box score

Grade B, low risk: 18 of 100 risk points, rated 2026-10-20. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.

This skill is an orchestration wrapper that fans out parallel subagents, gathers their results, and returns one report. It does not run shell code itself, but it does send task context to cloud workers and reads a model-selection dotfile under the user's home directory.

Category levels, from 0 (none) to 3 (high)
Category	Level	Why
Code & shell execution	0 (none)	No shell commands, code execution, or shipped scripts are present; it only describes agent orchestration.
Network access	2 (moderate)	It explicitly spawns workers with `environment: "cloud"` and `run_in_background: true`, which sends task context to remote subagents.
Filesystem reach	2 (moderate)	It reads the home dotfile `~/.cursor/rules/pstack-models.mdc` to choose a worker model, which reaches outside the project.
Secrets & credentials	0 (none)	It does not request tokens, passwords, `.env` files, keychains, or browser sessions.
Privilege & persistence	0 (none)	No sudo, service installation, shell startup edits, or other privilege changes are requested.
Instruction hijack surface	0 (none)	There are no hidden instructions, instruction-skip behaviors, or fetched external prompts.
Irreversible actions	1 (low)	Its purpose is to launch and coordinate N workers, but it only aggregates results and does not itself deploy, post, pay, or delete.
Transparency	0 (none)	All content is plain markdown; there are no binaries, blobs, or decoded/runtime-loaded code.

Before installing

Confirm you are okay with task context being sent to cloud subagents.
Verify `~/.cursor/rules/pstack-models.mdc` exists and contains the intended `swarm workers` line, or expect the fallback model.
If you plan to use `environment: "local"`, make sure the task does not expose sensitive local files.

Rated 2026-10-20 by an automated review of 1 file from the skill's repository.

Source

Repository: cursor/plugins (all skills from this repository)
Path: pstack/skills/swarm/SKILL.md
Branch: main
Collection: cursor
Updated: 2026-09-23

automate-me — Use for "automate me", "create/update/refresh my -mode skill", "turn/capture my preferences or working style into a skill".
continual-learning — Orchestrate continual learning by delegating transcript mining and AGENTS.md updates to `agents-memory-updater`.
orchestrate — Use only when the user explicitly types `/orchestrate <goal>` to decompose a large task.
principle-build-the-lever — Apply to any non-trivial work, not just bulk work: edits, migrations, analyses, checks.
principle-guard-the-context-window — Apply when context is filling up: large outputs, long files, repeated reads, fan-out planning.
ralph-loop — Start a Ralph Loop for iterative self-referential development.