project-indexing

# Project Indexing

## Overview

**Purpose**: Shannon's codebase compression system that achieves 94% token reduction (58K → 3K tokens) by generating structured SHANNON_INDEX files. Transforms linear file-by-file exploration into instant structured lookups, enabling fast agent onboarding, efficient multi-agent coordination, and sustainable context window usage.

**ROI Proven**: 40,000+ tokens saved per project analysis, 12-60x speedup, eliminates redundant file reads in multi-agent scenarios.

## When to Use

Use this skill when:
- Starting ANY project analysis or implementation (always generate index first)
- Onboarding new agents to existing codebase
- Launching multi-agent wave execution
- Switching between multiple projects/codebases
- Context window efficiency is critical
- After major codebase changes (regenerate index)

DO NOT use when:
- **NEVER skip** - Even "small" or "focused" questions benefit from indexing (see Anti-Rationalization section)
- Index already generated and current (< 24 hours old, no major changes)

## Core Competencies

1. **94% Token Compression**: Reduces 58K full codebase to 3K structured summary through hierarchical summarization and pattern abstraction
2. **Project Scan**: Discovers file counts, languages, LOC, dependencies without loading full content (500 token cost)
3. **Architecture Summarization**: Identifies core modules, key patterns, tech stack from directory structure and metadata (1,500 token cost)
4. **Context Enrichment**: Adds git recent changes, dependency analysis, testing setup (500 token cost)
5. **Template Population**: Generates structured SHANNON_INDEX.md following 7-section template (500 token cost)
6. **Serena Persistence**: Stores index in memory for cross-session retrieval and wave agent coordination (100 token cost)
7. **Multi-Agent Optimization**: Enables 81-95% token savings in parallel wave execution scenarios

## Inputs

**Required:**
- `project_path` (string): Absolute path to project root directory

**O