turbo-index

name	turbo-index
description	Index the current project for optimized search with QMD semantic search and fast file suggestions. Run this when entering a new codebase or after significant changes. Saves 60-80% tokens on exploration tasks.

Turbo Index

You are running the turbo-index skill to set up optimized search for this project.

Instructions

Follow these phases in order. Use the Bash tool to run commands. Report progress to the user after each phase.

Phase 1: Check Dependencies

Run the dependency checker:

~/.claude/plugins/*/claude-turbo-search/scripts/install-deps.sh --check-only 2>/dev/null || \
  ~/claude-turbo-search/scripts/install-deps.sh --check-only 2>/dev/null || \
  echo "DEPS_SCRIPT_NOT_FOUND"

If dependencies are missing, ask the user if they want to install them. If yes, run without --check-only:

~/.claude/plugins/*/claude-turbo-search/scripts/install-deps.sh 2>/dev/null || \
  ~/claude-turbo-search/scripts/install-deps.sh 2>/dev/null

Phase 2: Global Setup (if needed)

Check if file-suggestion.sh exists in ~/.claude/:

[ -f ~/.claude/file-suggestion.sh ] && echo "FILE_SUGGESTION_EXISTS" || echo "FILE_SUGGESTION_MISSING"

If missing, run the setup scripts:

# Setup file suggestion
~/.claude/plugins/*/claude-turbo-search/scripts/setup-file-suggestion.sh 2>/dev/null || \
  ~/claude-turbo-search/scripts/setup-file-suggestion.sh 2>/dev/null

# Setup QMD MCP server
~/.claude/plugins/*/claude-turbo-search/scripts/setup-mcp.sh 2>/dev/null || \
  ~/claude-turbo-search/scripts/setup-mcp.sh 2>/dev/null

Phase 3: Run Cartographer (if needed)

Check if codebase map exists and its age:

if [ -f docs/CODEBASE_MAP.md ]; then
  # Check if older than 24 hours (86400 seconds)
  AGE=$(($(date +%s) - $(stat -f %m docs/CODEBASE_MAP.md 2>/dev/null || stat -c %Y docs/CODEBASE_MAP.md 2>/dev/null)))
  if [ $AGE -gt 86400 ]; then
    echo "CODEBASE_MAP_STALE"
  else
    echo "CODEBASE_MAP_FRESH"
  fi
else
  echo "CODEBASE_MAP_MISSING"
fi

If missing or stale, use the Skill tool to invoke cartographer:

Skill: cartographer

Phase 4: Index with QMD

Get the project name and create/update the QMD collection:

PROJECT_NAME=$(basename "$PWD")
echo "Indexing project: $PROJECT_NAME"

# Add the project directory as a QMD collection
# Using correct QMD syntax: qmd collection add [path] --name <name> --mask <pattern>
qmd collection add . --name "$PROJECT_NAME" --mask "**/*.md"

# Add context for the collection
qmd context add . "Codebase documentation and structure for $PROJECT_NAME"

# Create vector embeddings for semantic search
echo "Creating vector embeddings (this may take a moment on first run)..."
qmd embed

Store metadata for staleness detection:

PROJECT_NAME=$(basename "$PWD")
mkdir -p .claude
cat > .claude/turbo-search.json << EOF
{
  "project": "$PROJECT_NAME",
  "lastIndexed": "$(date -u +%Y-%m-%dT%H:%M:%SZ)",
  "gitCommit": "$(git rev-parse HEAD 2>/dev/null || echo 'not-a-git-repo')",
  "qmdCollection": "$PROJECT_NAME"
}
EOF
echo "Metadata saved to .claude/turbo-search.json"

Phase 5: Report Results

Get indexing stats and report to user:

PROJECT_NAME=$(basename "$PWD")

# Common exclusions for dependencies/build artifacts across languages
EXCLUDES=(
  # Node.js/JavaScript
  --glob '!node_modules/' --glob '!bower_components/' --glob '!.npm/'
  --glob '!.yarn/' --glob '!.pnpm-store/' --glob '!dist/' --glob '!build/' --glob '!coverage/'
  # Elixir
  --glob '!_build/' --glob '!deps/' --glob '!.elixir_ls/' --glob '!.fetch/'
  # Java
  --glob '!target/' --glob '!.gradle/' --glob '!.mvn/' --glob '!out/' --glob '!.settings/'
  # Ruby
  --glob '!vendor/' --glob '!.bundle/' --glob '!.gem/'
  # PHP
  --glob '!vendor/' --glob '!.composer/'
  # Python (bonus)
  --glob '!__pycache__/' --glob '!.venv/' --glob '!venv/' --glob '!.tox/'
  --glob '!.mypy_cache/' --glob '!.pytest_cache/' --glob '!*.egg-info/'
  # General
  --glob '!.git/'
)

# Count indexed files (excluding dependencies)
TOTAL_FILES=$(rg --files --follow --hidden "${EXCLUDES[@]}" . 2>/dev/null | wc -l | tr -d ' ')
MD_FILES=$(rg --files --glob   . 2>/dev/null |  -l |  -d )

 
 
 
 
 
 
 
 
 
 
 
 
 

Using QMD for Token Savings

After indexing, always prefer searching before reading:

# Find relevant files first (fast, low tokens)
qmd search "your topic" --files -n 5

# Then read only what you need
qmd get "path/to/relevant-file.md"

The /qmd skill is also available - use it when you need to search for content across your indexed projects.

Notes

• This skill is idempotent - safe to run multiple times
• First run installs dependencies and configures Claude Code globally
• Subsequent runs just refresh the project index
• Restart Claude Code after first run to activate MCP tools
• Use qmd search before reading files to save tokens