Back to list
aiskillstore

spawn-parallel

by aiskillstore

Security-audited skills for Claude, Codex & Claude Code. One-click install, quality verified.

102🍴 3📅 Jan 23, 2026

SKILL.md


name: spawn-parallel description: Pattern for spawning parallel subagents efficiently. Use when you need multiple independent tasks done concurrently. allowed-tools: Task

Spawn Parallel Skill

Pattern for spawning and coordinating parallel subagents.

When to Load This Skill

  • You have multiple independent tasks
  • Tasks don't depend on each other's output
  • You want to maximize concurrency

Spawning Pattern

1. Identify Independent Tasks

Tasks are independent if:

  • No data dependencies between them
  • No file conflicts (different files or read-only)
  • Can complete in any order

2. Prepare Contexts

Each subagent needs minimal, focused context:

{"task":{"id":"unique_id","description":"specific task"},"context_files":["only relevant files"],"boundaries":{"owns":["files this agent can modify"],"reads":["files for reference"]},"output_path":"memory/tasks/{id}/output.json"}

3. Spawn All at Once

Use multiple Task calls in single response:

Task(subagent_type: "implementer", model: "sonnet", prompt: "Task 1...")
Task(subagent_type: "implementer", model: "sonnet", prompt: "Task 2...")
Task(subagent_type: "implementer", model: "sonnet", prompt: "Task 3...")

Subagent Type Reference (Custom Dotagent Agents):

TypeModelUse For
explorerhaikuFast codebase scouting
implementersonnetFocused code writing
verifierhaikuIndependent verification
testerhaikuTest execution

Note: These are custom dotagent agents (lowercase). Built-in Claude Code agents like Explore and Plan (capitalized) have different behavior.

4. Collect and Validate

After all complete:

  • Check each output file exists
  • Validate against schema
  • Handle failures (retry or escalate)

Coordination Rules

Prevent Conflicts

  • Define clear file ownership per agent
  • Use contracts for shared interfaces
  • Read-only access to shared resources

Handle Failures

Individual failures don't fail the batch. Apply recovery strategies from @.claude/skills/error-recovery/SKILL.md:

FOR each failed task in batch:
  IF output malformed/timeout:
    → Simple Retry (same prompt, up to 3x)
  ELIF agent said "unclear"/"don't understand":
    → Context Enhancement (add files, clarify)
  ELIF partial completion:
    → Scope Reduction (split into subtasks)
  ELIF boundary/contract violation:
    → Escalation (spawn contract-resolver)
  ELIF 3+ attempts failed:
    → Abort, record failure, continue with others

Retry with Context Enhancement Example:

Task(
  subagent_type: "implementer",
  model: "sonnet",
  prompt: |
    ## RETRY - Previous attempt failed
    Error: "Unclear how to connect to database"

    ## Additional Context
    See database config: @src/config/database.ts
    Connection pattern: @src/services/db-connection.ts

    ## Original Task
    {original_task_description}

    Output: memory/tasks/{id}/output.json
)

Example: Parallel Explorers

# Spawn 3 custom explorer agents in parallel
Task(
  subagent_type: "explorer",  # Custom dotagent agent
  model: "haiku",
  prompt: "Explore authentication code. Return compact JSON with findings."
)
Task(
  subagent_type: "explorer",
  model: "haiku",
  prompt: "Explore API routes. Return compact JSON with findings."
)
Task(
  subagent_type: "explorer",
  model: "haiku",
  prompt: "Explore database models. Return compact JSON with findings."
)

All run concurrently, results collected when all complete.

Example: Mixed Agent Types

# Parallel implementation with different boundaries
Task(
  subagent_type: "implementer",
  model: "sonnet",
  prompt: |
    Task: Add user validation
    Boundaries: owns=[src/validators/user.ts], reads=[src/types/]
    Output: memory/tasks/task-001/output.json
)
Task(
  subagent_type: "implementer",
  model: "sonnet",
  prompt: |
    Task: Add email service
    Boundaries: owns=[src/services/email.ts], reads=[src/config/]
    Output: memory/tasks/task-002/output.json
)

Score

Total Score

60/100

Based on repository quality metrics

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

0/10
説明文

100文字以上の説明がある

0/10
人気

GitHub Stars 100以上

+5
最近の活動

1ヶ月以内に更新

+10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

+5

Reviews

💬

Reviews coming soon