---

name: vanguard description: Scout swarm for broad domain mapping and edge case discovery. Launches multiple high-temperature agents with distinct search vectors. Use for finding obscure data, edge cases, and mapping new territories.

The Vanguard: Scout Swarm Protocol

When to Activate

Use this skill when the user requests:

• Mapping a new domain broadly
• Finding edge cases and outliers
• Searching for obscure data points
• Wildcard/contrarian searches
• User explicitly mentions "vanguard", "scout", or "broad search"

Architecture

10 Parallel Scouts with Distinct Search Vectors

The Swarm

1. Scatter - Launch 10 agents with high temperature (creative/divergent)
2. Direct - Each scout has a distinct search vector:
   • Scout 1: "Assume the consensus is wrong"
   • Scout 2: "Look for failures and negative examples only"
   • Scout 3: "Find the contrarian experts"
   • Scout 4: "Search non-English/non-Western sources"
   • Scout 5: "Look for historical parallels"
   • Scout 6: "Find the edge cases"
   • Scout 7: "What would a critic say?"
   • Scout 8: "Search academic/technical sources"
   • Scout 9: "Find the recent developments (last 6 months)"
   • Scout 10: "What's everyone missing?"

The Gathering

Returns a "Menu of Possibilities" - deliberately NOT synthesized to preserve outliers

Invocation

Copypython3 ~/.claude/skills/vanguard/vanguard.py "Your domain to scout"

Options

Copypython3 ~/.claude/skills/vanguard/vanguard.py "Your query" --scouts 5  # Fewer scouts for faster results

Example Usage

User: "What are the risks of Australia's green steel strategy?"

Copypython3 ~/.claude/skills/vanguard/vanguard.py "What risks and failure modes exist for Australia's green steel ambitions?"

Why This Works

1. Deliberate Divergence: High temperature prevents convergence
2. Distinct Vectors: Each scout looks in a different direction
3. Preserved Outliers: No synthesis means edge cases aren't averaged away
4. Contrarian by Design: Multiple scouts explicitly seek counter-evidence

Requirements

• Claude Code CLI installed and authenticated (claude command available)
• Python package: rich (for formatted output)