Skip to main content

swarm

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

소스 정보

저장소
Aqua-123/pstack-for-codex
최근 소스 활동
2026년 10월 5일 13:35
감지된 SKILL.md 언어
영어
스타
92
포크
16

설치 방법

기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.

소스 파일 검토

설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.

파일 탐색기
2 개 파일

SKILL.md 표시 중

SKILL.md
소스 지침 · 읽기 전용 미리보기
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.
# Swarm Delegation and optional capabilities follow `../poteto-mode/references/codex-agent-runtime.md`. Fan out N isolated workers. They may cover separate slices, race the same brief, or mix both. The parent waits, aggregates, and returns one report. If subagents or safe isolation are unavailable, use the declared sequential-parent or partial-result fallback. ## Start Open a todolist 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. Set N from the user or derive it from the shape. Cap N at observable runtime capacity. Queue excess lanes; never drop them silently. 4. Pick the worker model from `swarm workers` in the installed pstack model profiles when present. Otherwise use the configured fast profile. For a model race, name each arm's model up front. 5. Give each writer its own writable output under the runtime contract. When a worker verifies or measures commits, name the exact SHAs in its brief. Measurement briefs also name the sample count, what one sample means, and run order. The worker records the SHAs and method in its result. ## Phase B: Fan out After proving independent reads or isolated writes, dispatch all ready workers together through supported subagent tools. Use an installed role profile when validated; otherwise use generic agents with inherited model pairs. A worker that needs local devices or live-control capabilities stays on the surface that provides them. For a non-default base, create or select the exact worktree or branch before dispatch and name it in the brief. 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, reconcile its partial state and make one bounded retry when safe. Otherwise proceed with a labeled partial result or fail closed when that lane is required. ## 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. 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.
GitHub에서 보기