Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Converts rough research questions into complete USACF multi-agent swarm configurations. Runs a short interview to collect research parameters, auto-selects an algorithm (CoT, ToT, or GoT) based on complexity, then outputs all phase agents, memory namespace operations, and claude-flow commands. Includes adversarial red-team and fact-checker agents in every complex configuration.
Automates multi-agent research setup end-to-end, including adversarial review and algorithm selection, which manual prompting consistently skips.
When to use it
- Turning a vague competitive analysis prompt into a structured multi-agent research plan
- Investigating a new technology before committing to adoption
- Running due diligence on a software product or startup
- Converting a voice-dictated research idea into an executable swarm config
- Generating a gap analysis with built-in adversarial review across multiple domains
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: research what Solana is doing with AI and how we compare
- 1Detect medium complexity (comparison topic) and select Tree-of-Thought with 8 agents
- 2Run smart interview to collect research title, subject type, objectives, and output format
- 3Generate Phase 0 initialization commands using claude-flow swarm init with hierarchical topology
- 4Spawn parallel discovery and analysis agents, each writing to namespaced memory
- 5Add adversarial phase with red-team-reviewer and fact-checker, then produce final executive brief with quality score
Complete USACF super-prompt with claude-flow commands, 8 named agents, memory namespace mappings, and a quality score comparison (e.g. 1.2/10 → 9.3/10)
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.