# Parallel Agents

> Fan out independent read and write lanes.

- Skill: `darknesskiller/parallel-agents` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add darknesskiller/parallel-agents`
- Raw SKILL.md: https://api.skillmd.com/api/skills/darknesskiller/parallel-agents/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: darknesskiller (https://skillmd.com/u/darknesskiller)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/darknesskiller/parallel-agents

---


# Parallel Agents

User-invoked fan-out. Use when user asks for parallel or multi-agent work.

Fan out with the `workflow` tool. One run owns all lanes.
Set `model` and `effort` per lane with `agent({ model, effort })`.

Small fan-outs: `Agent` tool works too. Launch each lane, integrate in main thread.

Never one agent per one-line lookup.

## Steps

1. **Slice lanes.** Read-only first: one agent per independent question, directory, package, axis. Launch concurrently.

2. **Prompt tightly.** Give each agent: lane, read/write mode, allowed paths, forbidden paths, expected output shape.

3. **Write only when non-overlapping.** Fixed-contract, non-overlapping file sets. Two lanes might touch same file: keep sequential.

4. **Keep moving.** While agents run: do local reading, planning, small safe edits. Do not poll.

5. **Integrate.** Read summaries. For writers: verify diffs. Resolve seams in main thread. Run narrowest useful check.

## Rules

No duplicate lanes. No agents for one-line lookups. Never delegate ownership. Prefer more read agents over more write agents.