parallel-exec

自动任务分解和并行代理执行,适用于分解复杂任务、并行代理工作、多文件更改、优化执行时间、任务分解或理解并行执行阶段。

📖 中文导读:自动任务分解和并行代理执行,适用于分解复杂任务、并行代理工作、多文件更改、优化执行时间、任务分解或理解并行执行阶段。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: parallel-exec description: Automatic task decomposition with parallel agent execution. Use when: breaking down complex tasks, parallel agent work, multi-file changes, optimizing execution time, task decomposition, understanding parallel execution phases.

Parallel Exec

Automatically decomposes tasks and executes with parallel agents.

Usage

Skill(skill="parallel-exec", args="Implement login feature with UI, backend, and tests")

What It Does

  1. Analyzes the task description for parallelizable work items
  2. Decomposes into independent subtasks
  3. Builds a dependency graph
  4. Identifies parallelizable groups
  5. Executes with optimal parallelism
  6. Aggregates results with timing/metrics

Options

Pass options as a JSON string or comma-separated:

Skill(skill="parallel-exec", args='{"task": "Implement X", "parallel": true, "dry_run": false}')

Or simple form:

Skill(skill="parallel-exec", args="Implement login feature")

Output

## Decomposition Result

Original Task: [task description]

### Subtasks (N)
1. [EXPLORE] Explore codebase patterns
2. [ARCHITECT] Design architecture
3. [BUILDER] Implement backend logic
4. [UI] Design login dialog
5. [QUALITY] Write tests
6. [REVIEWER] Code review

### Execution Plan
- Phase 1 (Parallel): subtask_1, subtask_2, subtask_3
- Phase 2 (Sequential): subtask_4
- Phase 3 (Parallel): subtask_5, subtask_6

### Results
- Status: completed/partial/failed
- Total time: X ms
- Parallel groups: N
- Estimated savings: X ms

Parallel Opportunities

Task Pattern Detected Work Parallel Phases
"Implement X with Y and Z" 3 components Explore (×3) → Architect → Implement (×3) → Tests + Review
"Add feature for layer A and B" 2 layers Explore A + B → Implement A + B
"Fix bug in X and add tests" Bug + Tests Explore → Fix → Tests (sequential)

Integration

Uses the unified ChainExecutor with parallel mode:

from casare_rpa.domain.services import ChainExecutor

executor = ChainExecutor(orchestrator, enable_parallel=True)
result = await executor.execute("Implement login feature", parallel=True)