skill-parallel-agents

Orchestrates parallel multi-provider AI workflows for research, development, review, and delivery.

Updated Aug 15, 2026
One-click install
npx skills add https://github.com/GongYuanCaiJi/dsh-claude-octopus --skill skill-parallel-agents-gongyuancaiji
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: skill-parallel-agents
Source: https://github.com/GongYuanCaiJi/dsh-claude-octopus/tree/main/.claude/skills/skill-parallel-agents
Command: npx skills add https://github.com/GongYuanCaiJi/dsh-claude-octopus --skill skill-parallel-agents-gongyuancaiji

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Single-model AI assistance can miss perspectives, introduce blind spots, and lack validation on complex tasks. This Skill decomposes large tasks across multiple AI providers (Codex, Antigravity, Claude, OpenRouter, and others) running in parallel, then enforces consensus gates before delivery. ## Core Features & Use Cases - Double Diamond Workflow: Four-phase methodology (probe, grasp, tangle, ink) covering Discover, Define, Develop, and Deliver with quality gates between phases. - Adversarial Cross-Model Review: Grapple mode makes providers critique each other's proposals; Squeeze mode runs red-team security reviews against implementations. - Smart Auto-Routing: Detects intent keywords (research, build, review, security audit) and routes to the appropriate workflow with cost-aware provider selection. - Use Case: Ask to research OAuth patterns, and multiple providers explore the topic in parallel, a 75% consensus gate filters disagreements, and you receive a synthesized structured report. ## Quick Start Ask the assistant to research a topic, build a feature, or review code for security issues, and the multi-provider orchestration will activate automatically.

Frequently Asked Questions about skill-parallel-agents

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

FAQPage Schema
How do I run a task with multiple AI providers in parallel?▼

Use natural language requests like research, build, or review, or force parallel mode with the /octo:multi command. The orchestrate.sh script dispatches the task to available providers such as Codex and Antigravity simultaneously and synthesizes the results.

What is the Double Diamond workflow in Claude Octopus?▼

It is a four-phase methodology: probe (Discover) researches the problem, grasp (Define) builds consensus on requirements, tangle (Develop) implements with quality gates, and ink (Deliver) validates the final output. The embrace command runs all four phases sequentially.

Do I need both Codex and Antigravity installed to use multi-provider orchestration?▼

No, one provider is sufficient. The detect-providers command checks availability, and workflows gracefully degrade to a single provider, using it multiple times for tasks that would normally span several providers.

How does adversarial cross-model review work for security?▼

The squeeze command runs a red-team review where one provider implements secure code and another attacks it to find vulnerabilities, followed by remediation and validation phases. Grapple mode similarly makes providers critique each other's proposals.

When should I not use parallel multi-provider execution?▼

Avoid it for built-in Claude Code commands, simple file operations, git commands, and basic terminal tasks, which should use standard tools directly. Forcing parallel mode also incurs external API costs of roughly $0.02-0.08 per query.

Why does the quality gate fail during the tangle phase?▼

The tangle phase requires a 75% success rate across parallel agents. If it fails, break the task into smaller subtasks, increase the timeout with the -t flag, or inspect individual agent logs in the workspace logs directory.