---
name: swarm
description: "Fan out N parallel workers, drain them, and return one report. Use for /swarm, 'swarm this', or parallel coverage, races, gauntlets, and exploration."
disable-model-invocation: true
---

# Swarm

In Codex, read [the Codex adapter](../../docs/harness/codex.md) once per session before you follow this skill. It maps the Claude Code names in pstack skills to Codex.

Another pstack skill named here, as the **name** skill or `/name`, lives at `../<name>/SKILL.md`. Read that file and follow it, because the `Skill` tool refuses most pstack skills. When you tell the user to type one, write `/pstack:<name>`.

## Start

Open a todo list with one entry per phase before launching anything.

1. Frame
2. Fan out
3. Aggregate
4. Report

## Phase A: Frame

1. State the done predicate and the artifact or report the swarm must return.
2. Choose the shape. Partition into slices, race N workers on identical briefs, or mix both. For a race or mixed shape, declare `first pass`, `rank all`, or `best-of` before spawning.
3. Pick the worker model from the `swarm workers` line in `~/.claude/rules/pstack-models.md`. If the rule or that line is missing, use `sonnet medium`. A value is `<model> <effort>`, passed as the `Agent` call's `model` and `effort`. For `inherit`, omit both so the workers run on the parent model. If the `Agent` tool rejects a value, use the default and say so. If it rejects the default too, follow the fallback in the Subagents section of the **poteto-mode** skill. For a model race, name each arm's model up front. Pick each subagent's effort for the job, per the Subagents section of the **poteto-mode** skill.
4. Give each worker its own writable output when it writes. When workers verify or measure commits, each brief names the exact SHAs. A measurement brief also names the method (sample count, what one sample is, order). The worker records both in its result.

## Phase B: Fan out

Spawn all N workers in one message with `subagent_type: "general-purpose"`, `isolation: "worktree"`, `run_in_background: true`, and the step 3 `model` and `effort`, left unset for `inherit`. Omit `isolation` only when the worker must see this checkout's uncommitted state.

When a worker must start from a non-default pushed branch, name the branch in its brief and tell it to check that branch out first.

Every brief stands alone. Include the goal, scope, exact slice or race arm, how to verify, and what to report. Reports use `PASS`, `ISSUES`, or `BLOCKED` with evidence. A worker that can prove a defect reports `ISSUES` and lists every issue it can prove, not only the first.

If a worker drops out, proceed with N-1 and note it.

## Phase C: Aggregate

Read the terminal results. Drop a result that does not record the SHAs and method its brief names, and respawn that worker once. After a second miss, record a gap. A gap does not count as a pass. For coverage, every required slice needs a result. For a race, apply the selection rule declared up front. Use first pass, rank all, or best-of. Do not paste raw worker dumps.

## Phase D: Report

Return one consolidated in-chat report with a result table, issue one-liners, gaps or dropouts, and the race rule when used.