swarm

Orchestrates parallel subagent exploration, reviewer validation, and critic challenge for session tasks.

Updated Mar 30, 2026
One-click install
npx skills add https://github.com/ZaxbyHub/ragappv3 --skill swarm-zaxbyhub
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: swarm
Source: https://github.com/ZaxbyHub/ragappv3/tree/main/.opencode/skills/swarm
Command: npx skills add https://github.com/ZaxbyHub/ragappv3 --skill swarm-zaxbyhub

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Complex coding tasks often ship with undetected defects because a single agent context both implements and verifies its own work, leading to shallow review and premature completion. ## Core Features & Use Cases - Swarm Mode Activation: Writes a session contract file that switches the agent into a swarm-like orchestration mode for the current session. - Layered Validation Ladder: Separates explorer (candidate generation), reviewer (independent validation), and critic (final challenge) roles across isolated subagent contexts. - Risk-Based Triage: Applies deep validation only to high-risk areas like auth, payments, migrations, and concurrency while keeping low-risk changes lightweight. - Use Case: Run /swarm refactor the auth refresh flow to enable swarm mode and immediately execute the refactor with parallel exploration, independent reviewer sign-off, and CI compatibility checks before pushing. ## Quick Start Ask the agent to run /swarm followed by your task, such as fixing failing auth tests, and it will enable swarm mode and execute the task with parallel subagents and independent validation.

Frequently Asked Questions about swarm

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I enable swarm mode for a coding session?▼

Run the /swarm command with no arguments to enable swarm mode for the current session. It writes a swarm-mode contract file to the session directory, and all subsequent complex tasks follow the parallel exploration and validation workflow automatically.

How do I run a task immediately with swarm orchestration?▼

Pass the task as an argument, for example /swarm fix the failing auth refresh tests. The skill enables swarm mode first, then executes the task through the full ladder: parallel exploration, scoped implementation, reviewer validation, and critic challenge when needed.

What is the difference between explorer, reviewer, and critic subagents?▼

Explorers map code broadly and produce candidate findings only. Reviewers independently validate candidates and classify them as CONFIRMED, DISPROVED, UNVERIFIED, or PRE_EXISTING. Critics challenge reviewer-confirmed findings in small batches to remove weak or overclaimed results.

Does swarm mode slow down simple or low-risk changes?▼

No. The skill uses risk-based triage: docs-only changes, small UI copy edits, and isolated cleanups follow a lightweight path, while high-risk work like auth, payments, migrations, and concurrency gets the full validation ladder.

Does swarm mode persist after the session ends?▼

No. Swarm mode only writes a session-scoped file and does not permanently change project behavior. Re-run /swarm after clearing or resetting the session context to re-enable it.