Habilidades super-swarm-spark

super-swarm-spark

Riesgo bajo ⚙️ Comandos externos

Run 12 Parallel AI Agents on Your Dev Plan

Executing large development plans sequentially is slow. This skill orchestrates up to 12 concurrent AI subagents that work in parallel on different tasks, dramatically speeding up project completion while maintaining coordination.

Soporta: Claude Codex Code(CC)
⚠️ 65 Deficiente
1

Descargar el ZIP de la skill

2

Subir en Claude

Ve a Configuración → Capacidades → Skills → Subir skill

3

Activa y empieza a usar

Pruébalo

Usando "super-swarm-spark". /super-swarm-spark auth-plan.md

Resultado esperado:

Starting parallel execution with 12 Sparky agents...

Launched: T1 (User Model), T2 (Auth Controller), T3 (JWT Handler)...

Task T1 completed: Created User model with validation
Task T2 completed: Implemented auth endpoints

Integration pass: Resolved 2 path conflicts

Final status: 8/8 tasks completed, all tests passing

Usando "super-swarm-spark". /super-swarm-spark ./plans/api-plan.md T1 T2 T4

Resultado esperado:

Running subset: T1, T2, T4

Launched 3 concurrent agents...

All tasks completed successfully.
Updated plan file with completion logs.

Auditoría de seguridad

Riesgo bajo
v1 • 3/7/2026

All 22 static findings are false positives. The scanner misinterpreted markdown backticks as shell execution commands and generated spurious crypto algorithm warnings. This skill is a legitimate agent orchestrator that manages parallel subagent execution for development tasks. No malicious behavior detected.

1
Archivos escaneados
219
Líneas analizadas
4
hallazgos
1
Auditorías totales

Problemas de riesgo alto (1)

False Positive: Weak Cryptographic Algorithm
Static scanner incorrectly flagged lines as containing weak cryptographic algorithms. Analysis shows NO cryptographic code exists in this skill. Lines flagged (3, 38, 44, 66, 93, 126) contain only orchestration instructions and task prompts.
Problemas de riesgo medio (1)
False Positive: External Commands (Markdown Backticks)
Static scanner detected 'Ruby/shell backtick execution' but these are markdown code formatting (backticks around words like `pending`, `running`). No actual shell execution occurs.
Problemas de riesgo bajo (1)
False Positive: System Reconnaissance
Scanner flagged line 71 mentioning 'launch' as system reconnaissance. This is standard agent orchestration behavior, not reconnaissance.
Auditado por: claude

Puntuación de calidad

38
Arquitectura
100
Mantenibilidad
87
Contenido
32
Comunidad
73
Seguridad
83
Cumplimiento de la especificación

Lo que puedes crear

Accelerate Large Feature Implementations

When implementing a large feature that can be broken into independent tasks, use this skill to run multiple subagents in parallel, reducing total execution time significantly.

Parallelize Test Writing

Run multiple subagents simultaneously to write tests for different modules or components of a codebase in parallel.

Batch Refactoring Tasks

Execute multiple refactoring tasks across different parts of the codebase concurrently while maintaining naming consistency through context packs.

Prueba estos prompts

Run Full Development Plan
Execute the plan file ./my-project-plan.md using super-swarm-spark. Run all tasks in the plan.
Run Specific Tasks
Execute tasks T1, T3, and T5 from ./backend-plan.md using super-swarm-spark.
Continue After Partial Completion
Continue executing ./feature-plan.md. The following tasks are already complete: T1, T2. Run the remaining pending tasks.
Integration and Testing Pass
After all subagents complete their tasks in ./full-plan.md, perform the final integration pass and run tests to ensure everything works together.

Mejores prácticas

  • Ensure plan files have clear task IDs (T1, T2, etc.) and complete acceptance criteria for each task
  • Provide canonical file paths in the plan to help the orchestrator prevent naming conflicts
  • Keep tasks relatively independent to maximize parallel execution benefits

Evitar

  • Do not use this skill for plans with heavy dependencies between tasks (subagents will block waiting on each other)
  • Avoid running this skill without a properly formatted plan file
  • Do not expect the orchestrator to handle tasks requiring the same file simultaneously

Preguntas frecuentes

What is the difference between this skill and other orchestration skills?
This skill uses a rolling pool of up to 12 concurrent Sparky subagents with explicit path coordination to prevent filename drift across parallel tasks.
Can I use this with agent types other than Sparky?
No. This skill explicitly requires the sparky agent role and will not work with other agent types.
How does the skill handle file name conflicts?
The orchestrator builds context packs with canonical file paths and enforces exact naming in all subagent prompts to prevent conflicts.
What happens if a subagent task fails?
The orchestrator validates each completion, detects failures, and can retry or escalate failed tasks before proceeding.
Do I need to push my commits after the skill runs?
No. The skill instructs subagents to commit but never push. You review and push changes after verification.
Can I run a subset of tasks from my plan?
Yes. Provide task IDs after the plan file (e.g., /super-swarm-spark plan.md T1 T3 T5) to run only those tasks.

Detalles del desarrollador

Estructura de archivos

📄 SKILL.md