# Subagent Batching

> Patterns for launching and managing parallel subagents efficiently.

- Skill: `majiayu000/subagent-batching` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds add majiayu000/subagent-batching`
- Raw SKILL.md: https://api.skillmd.com/api/skills/majiayu000/subagent-batching/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: majiayu000 (https://skillmd.com/u/majiayu000)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/majiayu000/subagent-batching

---


# Subagent Batching

Patterns for parallel subagent execution. Use your judgment on batch sizes based on task complexity.

## When to Parallelize

**Parallelize when**:

- Processing multiple independent items (PRs, issues, files)
- Tasks have no dependencies between items
- Each item can succeed or fail independently

**Don't parallelize when**:

- Later items depend on earlier results
- Shared resources could cause conflicts
- You need to aggregate results before proceeding

## Launch Pattern

Launch subagents in a **single message** with multiple Task tool calls. Don't launch sequentially.

```text
# Correct: ONE message with multiple Task calls
Task 1: Process item A
Task 2: Process item B
Task 3: Process item C
# All launch simultaneously
```

## Batching for Large Sets

For many items, batch them and validate between batches:

1. Launch a batch of subagents in parallel (single message)
2. Wait for all to complete using `TaskOutput` with `block=true`
3. Validate results before proceeding
4. Start next batch after validation

## Subagent Prompts

Give each subagent focused context:

- Clear task description
- Relevant identifiers and locations
- Expected output format
- Success criteria

## Error Handling

- One failure shouldn't block the rest
- Mark failed items for human review
- Continue processing remaining items
- Provide summary of successes and failures

## Reporting

After completion, summarize:

- Total items processed
- Successes with brief descriptions
- Failures with reasons
- Any required follow-up actions

