widescreen-research: Command for Claude Code

Docs Verification Game

Game-theoretic documentation verification workflow that systematically validates accuracy, prevents false positives, and ensures documentation matches implementation reality using multi-dimensional testing and validation gates.

Command Structure

/docs-verification-game [DOCS_PATH] [CODEBASE_PATH] [--SCOPE=<scope>] [--DEPTH=<depth>] [--STRATEGY=<strategy>] [--CONFIDENCE=<threshold>]

Parameter Definitions

• DOCS_PATH: Location of documentation to verify (file, directory, or pattern)
• CODEBASE_PATH: Location of the codebase the docs describe
• SCOPE: Verification scope (api, examples, concepts, all) - default: all
• DEPTH: Verification depth (surface, standard, deep) - default: standard
• STRATEGY: Verification approach (sequential, parallel, adversarial) - default: parallel
• CONFIDENCE: Required confidence threshold (0.0-1.0) - default: 0.9

Game Variables

DOCS_PATH: $ARGUMENTS CODEBASE_PATH: $ARGUMENTS SCOPE: $ARGUMENTS DEPTH: $ARGUMENTS STRATEGY: $ARGUMENTS CONFIDENCE: $ARGUMENTS

Phase 1: Documentation Analysis & Inventory (Time-boxed: 25% of budget)

Objective: Map documentation claims and create verifiable assertions

# Initialize verification game state
echo "🔍 Starting Docs Verification Game"
echo "📄 Target: $DOCS_PATH"
echo "💻 Codebase: $CODEBASE_PATH"
echo "🎯 Required confidence: $CONFIDENCE"

mkdir -p .docs-verify/{state,analysis,assertions,evidence}
cat > .docs-verify/state.json << 'EOF'
{
  "phase": "analysis",
  "confidence_level": 0.0,
  "verification_rounds": 0,
  "false_positive_detections": [],
  "accuracy_issues": [],
  "players": {
    "skeptic": { "doubt_level": 0.9, "weight": 1.0 },
    "implementer": { "reality_check": 0.8, "weight": 0.9 },
    "user": { "confusion_potential": 0.5, "weight": 0.8 },
    "maintainer": { "drift_concern": 0.7, "weight": 0.7 }
  }
}
EOF

Step 1.1: Extract Documentation Claims

# Parse documentation for verifiable claims
echo "📊 Extracting verifiable claims from documentation..."

# Use Claude Code's semantic understanding to identify claims
/codebase-search "What specific APIs, methods, or behaviors are documented in $DOCS_PATH?" --target-directories "$DOCS_PATH"

# Categorize claims by type
cat > .docs-verify/analysis/claim-categories.json << 'EOF'
{
  "api_signatures": [],      // Function signatures, parameters, returns
  "code_examples": [],       // Executable code snippets
  "behavioral_claims": [],   // How things work, side effects
  "configuration": [],       // Config options, defaults
  "error_scenarios": [],     // Error handling, edge cases
  "performance_claims": [],  // Speed, memory, scalability
  "compatibility": []        // Version requirements, dependencies
}
EOF

# Extract and categorize each claim
find "$DOCS_PATH" -type f \( -name "*.md" -o -name "*.mdx" \) | while read -r doc_file; do
  echo "Analyzing: $doc_file"

  # Extract code blocks
  grep -n '```' "$doc_file" | awk 'NR%2==1 {start=$1} NR%2==0 {print start "-" $1}' > .docs-verify/analysis/code-blocks-$doc_file.txt

  # Extract API references
  grep -E "(function|class|interface|type|export)" "$doc_file" > .docs-verify/analysis/api-refs-$doc_file.txt
done