Swarm
Fan out N parallel workers. They may cover separate slices, race the same brief, or mix both. The parent waits, aggregates, and returns one report.
Start
Open a todolist with one entry per phase before launching anything.
- Frame
- Fan out
- Aggregate
- Report
Phase A: Frame
- State the done predicate and the artifact or report the swarm must return.
- 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.
- Set N from the user or derive it from the shape. N is total workers.
- Pick the worker model from the
swarm workers line in p3-models.md, resolved through orchestrator_capabilities. If the file or that line is missing, run setup-p3. For auto, inherit, or inherit-parent, omit target so the workers run on the parent model. If delegate_task rejects the target, fall back to the closest valid model of the same family from orchestrator_capabilities and say so. For a model race, name each arm's model up front.
- 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 turn with delegate_task, mode: "async", and the step 4 target, left unset for auto, inherit, or inherit-parent. Use t3_thread_launch with a workspaceStrategy only when a worker needs its own thread, worktree, or branch. One writer per worktree.
Every brief stands alone. Child agents get only the brief, never the parent's context. 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
Drain with task_status until every worker is terminal. 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.
Keep a compact result table, one-line evidenced issues, and explicit gaps or dropouts.
Phase D: Report
Return one consolidated in-chat report with the table, issue one-liners, gaps or dropouts, and the race rule when used.
This repository (Asymmetric-al/core)
Vendored from p3-stack (MIT) at commit 09909eb. Provenance is in references/upstream.md. On conflict, the current user request, root AGENTS.md, the nearest nested AGENTS.md, and the matching docs/ai/rules/* rulebook win over this skill.
- Explicit invocation only: use it when the user invokes
/swarm (Claude Code, Cursor, T3 Code) or $swarm (Codex), or when an invoked p3-stack skill routes here. Do not route other work to it.
- The T3 Code tools named above map to Claude Code, Codex, and Cursor in
docs/ai/skills/p3-mode/references/core-adapter.md. Read it before you delegate, watch a PR, schedule work, or drive a preview outside T3 Code.
- Confirm before outward-facing or hard-to-reverse actions such as pushing, merging, deploying, deleting, or posting messages, unless the user authorized that action for this task.
Edit this canonical copy in docs/ai/skills/swarm/, then run bun run skills:sync and bun run skills:verify.
1---2name: swarm3description: "Fan out N parallel workers, drain them, and return one report. Use for /swarm, 'swarm this', or parallel coverage, races, gauntlets, and exploration."4disable-model-invocation: true5metadata:6 source: https://skillmd.com/skills/asymmetric-al/swarm7---89# Swarm1011Fan out N parallel workers. They may cover separate slices, race the same brief, or mix both. The parent waits, aggregates, and returns one report.1213## Start1415Open a todolist with one entry per phase before launching anything.16171. Frame182. Fan out193. Aggregate204. Report2122## Phase A: Frame23241. State the done predicate and the artifact or report the swarm must return.252. 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.263. Set N from the user or derive it from the shape. N is total workers.274. Pick the worker model from the `swarm workers` line in `p3-models.md`, resolved through `orchestrator_capabilities`. If the file or that line is missing, run `setup-p3`. For `auto`, `inherit`, or `inherit-parent`, omit `target` so the workers run on the parent model. If `delegate_task` rejects the target, fall back to the closest valid model of the same family from `orchestrator_capabilities` and say so. For a model race, name each arm's model up front.285. 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.2930## Phase B: Fan out3132Spawn all N workers in one turn with `delegate_task`, `mode: "async"`, and the step 4 target, left unset for `auto`, `inherit`, or `inherit-parent`. Use `t3_thread_launch` with a `workspaceStrategy` only when a worker needs its own thread, worktree, or branch. One writer per worktree.3334Every brief stands alone. Child agents get only the brief, never the parent's context. 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.3536If a worker drops out, proceed with N-1 and note it.3738## Phase C: Aggregate3940Drain with `task_status` until every worker is terminal. 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.4142Keep a compact result table, one-line evidenced issues, and explicit gaps or dropouts.4344## Phase D: Report4546Return one consolidated in-chat report with the table, issue one-liners, gaps or dropouts, and the race rule when used.4748<!-- CORE-OVERLAY-START -->4950## This repository (Asymmetric-al/core)5152Vendored from [p3-stack](https://github.com/uzairansaruzi/p3-stack) (MIT) at commit `09909eb`. Provenance is in `references/upstream.md`. On conflict, the current user request, root `AGENTS.md`, the nearest nested `AGENTS.md`, and the matching `docs/ai/rules/*` rulebook win over this skill.5354- Explicit invocation only: use it when the user invokes `/swarm` (Claude Code, Cursor, T3 Code) or `$swarm` (Codex), or when an invoked p3-stack skill routes here. Do not route other work to it.55- The T3 Code tools named above map to Claude Code, Codex, and Cursor in `docs/ai/skills/p3-mode/references/core-adapter.md`. Read it before you delegate, watch a PR, schedule work, or drive a preview outside T3 Code.56- Confirm before outward-facing or hard-to-reverse actions such as pushing, merging, deploying, deleting, or posting messages, unless the user authorized that action for this task.5758Edit this canonical copy in `docs/ai/skills/swarm/`, then run `bun run skills:sync` and `bun run skills:verify`.5960<!-- CORE-OVERLAY-END -->