Przejdź do treści
Biblioteka skilli
AI i agenci · Automatyzacja

dispatching-parallel-agents

/dispatching-parallel-agents

Uruchamia równolegle dwa lub więcej niezależnych zadań, np. osobne pliki testów czy odrębne podsystemy.

Repozytorium
14All ↗
Licencja
MIT
Ostatnia zmiana repozytorium
05.11.2026

Ocena i bezpieczeństwo

100/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
2
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 05.11.2026, bez odwołań do wycofanych modeli i API.

Licencja

MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.

Co robi i kiedy

Uruchamia równolegle dwa lub więcej niezależnych zadań, np. osobne pliki testów czy odrębne podsystemy.

Oryginalny opis (EN)

Use when facing two or more independent tasks, such as unrelated failing test files or separate subsystems, that can be worked on without shared state or sequential dependencies

Jak działa

  1. When to use
  2. Inside an implementation plan
  3. The pattern
  4. What a good brief contains
  5. Common mistakes
  6. Example
Sekcje z nagłówków SKILL.md, w oryginale · 55 linii, 595 słów
Podgląd SKILL.md
---
name: dispatching-parallel-agents
description: Use when facing two or more independent tasks, such as unrelated failing test files or separate subsystems, that can be worked on without shared state or sequential dependencies
---

# Dispatching Parallel Agents

You delegate tasks to subagents with isolated context. Build exactly the instructions and context each one needs; never let them inherit your session history. This also preserves your own context for coordination.

When several problems are unrelated (different test files, different subsystems, different bugs), investigating them one after another wastes time. Dispatch one subagent per independent problem domain and let them work concurrently.

## When to use

Use it when:

- Three or more test files fail for different root causes.
- Several subsystems are broken independently.
- Each problem can be understood without context from the others.
- No shared state: the agents will not edit the same files or use the same resources.

Do not use it when:

- The failures are related and fixing one might fix others. Investigate together first.
- You need the full system state to understand the problem.
- You do not yet know what is broken (exploratory debugging).
- The agents would interfere with each other.

## Inside an implementation plan

When you execute a plan with the `subagent-driven-development` skill, its tasks run serial by default, one implementer after another. Use this skill for plan tasks only when they touch disjoint files and share no state or interfaces; otherwise keep them serial, because parallel edits to the same code conflict.

## The pattern

1. **Identify independent domains.** Group the work by what is broken. Each group must be fixable without touching another group's code.
2. **Write one focused brief per agent.** Each brief has a specific scope (one test file or subsystem), a clear goal, constraints ("do not change other code"), and the output you want back (a summary of the root cause and what changed).
3. **Dispatch all agents in the same response** so they run in parallel. Dispatching one per response runs them one after another.
4. **Review and integrate.** Read each summary, check that no two agents edited the same code, run the full test suite, and spot-check; agents can make systematic errors.

## What a good brief contains

- One clear problem domain.
- All the context needed: the failing test names and the error messages, pasted in.
- The method you want: understand what each test verifies, find the root cause, fix the cause and not the symptom. Say what not to do, such as "do not just raise the timeouts".
- The output format: root cause found, files changed, anything unresolved, and the files read with the line range actually seen. The file-reading tool truncates a long file, so the agent pages to the end or says where it stopped.
- Every input stated. An input you have no value for says "none provided" instead of being left out, so the agent does not invent one.

## Common mistakes

| Mistake | Fix |
|---|---|
| Too broad ("fix all the tests") | One test file or subsystem per agent |
| No context ("fix the race condition") | Paste the errors and test names |
| No constraints | Say what must not change, for example production code |
| Vague output ("fix it") | Ask for the root cause and a list of changes |

## Example

Six failures across three test files after a refactor: abort handling (timing), batch completion (events), and approval races (async wait). The domains are independent, so three agents are dispatched in one response. They return an event-based wait, a fixed event structure, and an added wait for async execution. The fixes do not conflict and the full suite passes.
Pliki w paczce (2)
  • SKILL.md 4 KB
  • meta.yaml 1 KB

Wersje

RepozytoriumWynikLicencja
Ta strona · LILLINH/14All100MIT
jjuraszek/pi-gauntlet100MIT
osmontero/opencode-skills100MIT
coco-research/coco100MIT
FrancoStino/opencode-skills-collection100MIT
seaworld008/commonly-used-high-value-skills100MIT
rootcastleco/rei-skills100MIT
nikospavlopoulos/devops-mutant-powers100MIT
Richicinschi/studio-os100MIT

Migawka bazy z 13.11.2026 · wersja paczki 94229049d9e1 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 100sr-dispatching-parallel-agents
    Dzieli dwa lub więcej niezależnych zadań między równoległych agentów, gdy nie dzielą stanu i nie zależą od siebie.
  • 100dispatch-multiple-agents
    Rozdziela dwa lub więcej niezależnych zadań bez wspólnego stanu i kolejności na subagentów, którzy pracują równolegle.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.