SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsparallel-execution

parallel-execution

Agents & models

View repository

Updated 10-15 20:30

About this skill

Parallel Execution

Execute multiple independent tasks simultaneously to maximize throughput and minimize total execution time.

Session Start Protocol: Identify parallelizable work and spawn task agents immediately at the start of a session, instead of doing everything sequentially.

When to Use

  • Multiple independent tasks (no dependencies)
  • Tasks benefit from concurrent execution
  • Available agents for parallel work
  • Results can be aggregated after completion

Core Concepts

Independence

Tasks are independent when:

  • ✓ No data dependencies
  • ✓ No resource conflicts
  • ✓ No ordering requirements
  • ✓ Failures are isolated

Concurrency

Critical: Use single message with multiple Task tool calls

Single message:
- Task → Agent A
- Task → Agent B
- Task → Agent C

All start simultaneously.

Synchronization

  • Wait for all agents to complete
  • Collect and validate results
  • Aggregate into final output

Process

Step 1: Identify Independent Tasks

Checklist:

  • No data dependencies
  • No shared writes
  • No execution order requirements
  • Failures don't cascade

Step 2: Agent Assignment

Available: code-reviewer, test-runner, feature-implementer, refactorer, debugger

Step 3: Launch Parallel Execution

Single message with multiple Task tool calls. All agents start simultaneously.

Step 4: Monitor Execution

  • Track progress every 2-3 minutes
  • Note completion times
  • Watch for errors

Step 5: Collect & Validate

As each completes:

  1. Collect output
  2. Validate against criteria
  3. Mark complete/failed

Step 6: Aggregate Results

## Results

1. ✓ Task A - Result
2. ✓ Task B - Result
3. ✓ Task C - Result

Status: ✓ Success

Execution Patterns

Homogeneous Parallel

Same agent type, different inputs:

├─ test-runner: Test module A
├─ test-runner: Test module B
└─ test-runner: Test module C

Heterogeneous Parallel

Different agent types:

├─ code-reviewer: Quality analysis
├─ test-runner: Test execution
└─ debugger: Performance profiling

Parallel with Convergence

Parallel execution → Single synthesis:

Phase 1: Parallel investigation
Phase 2: Synthesize findings

Synchronization Strategies

  • Wait for All (AND): All must complete
  • Wait for Any (OR): First success proceeds
  • Wait for Threshold: N of M must complete

Error Handling

Independent Failures

One failing doesn't stop others:

├─ Agent A: ✓ Success
├─ Agent B: ✗ Failed
└─ Agent C: ✓ Success

Collect A and C, report B failed

Strategies

  1. Fail Fast: Stop if any fails
  2. Best Effort: Collect successes
  3. Retry Failed: Retry failures

Performance

Speedup Calculation

Sequential = T1 + T2 + T3
Parallel = max(T1, T2, T3)
Speedup = Sequential / Parallel

Example: Tasks 10min, 15min, 8min

  • Sequential: 33 min
  • Parallel: 15 min (max)
  • Speedup: 2.2x

Best Practices

DO:

✓ Verify independence first ✓ Use single message with multiple tools ✓ Balance workload ✓ Handle failures gracefully ✓ Validate each result

DON'T:

✗ Parallelize dependent tasks ✗ Send sequential messages ✗ Overload single agent ✗ Skip validation

Examples

Simple Parallel Review

├─ code-reviewer: Review code
└─ test-runner: Run tests
Speedup: 2x

Multi-Module Testing

├─ test-runner: Test module A
├─ test-runner: Test module B
└─ test-runner: Test module C
Speedup: 3x

Quality Check

├─ code-reviewer: Quality
├─ test-runner: Tests
├─ test-runner: Benchmarks
└─ debugger: Memory check
Speedup: 4x

Integration

Used by agent-coordination for independent task strategy.

Summary

Parallel execution maximizes efficiency through concurrent execution, independent validation, and synchronized aggregation.

Reference Files