SantiagoPrada2005

Codebase Index — Security skill for Claude Code

Security community

Index-First Navigation for AI Agents: Reduce token consumption by 85% with deterministic directory indexing and zero-drift CI audit.

How to install Codebase Index

This entry records only its repository, not the path inside it, so there is no exact command to give. Open SantiagoPrada2005/codebase-index and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Codebase Index does

Index-First Navigation for AI Agents: Reduce token consumption by 85% with deterministic directory indexing and zero-drift CI audit.

Alternatives in Security

Cross-Pollinate Hex MCP Servers — Diff-driven audit: find real transferable changes across hex-line-mcp, hex-ssh-mcp, hex-graph-mcp 238 ★
Node9 Proxy — The Execution Security Layer for the Agentic Era 209 ★
Loongsuite Pilot — Local-first telemetry collector for AI coding agents — unified OpenTelemetry events for Claude Code, Codex, Cu 150 ★

README

🗺️ Codebase Index (`codebase-index`)

**Index-First Navigation for AI Agents**: Slash exploration tokens by up to **85%–90%**, eliminate hallucinated imports, and boost LLM reasoning quality by keeping context windows clean and noise-free.

[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT) [![AI Agent Compatible](https://img.shields.io/badge/AI%20Agent-Claude%20Code%20%7C%20Cursor%20%7C%20Antigravity-emerald)](https://github.com/SantiagoPrada2005/codebase-index) [![CI Workflow](https://img.shields.io/badge/CI-Zero--Drift%20Audit-purple)](.github/workflows/index-audit.yml) [![skills.sh Ready](https://img.shields.io/badge/skills.sh-Available-orange)](https://skills.sh)

⚡ The Problem: The "Agent Wandering" Tax & Context Dilution

Modern coding agents (Claude Code, Cursor, Antigravity, Windsurf, Codex) are powerful, but when placed in mid-to-large codebases without structured maps, they suffer from three structural failure modes:

❌ TRADITIONAL BLIND TRAVERSAL
Agent Prompt ──▶ Global grep / find ──▶ Reads 8–15 Full Files ──▶ Context Window Saturation
                                        (~12,000 tokens)          ├─ Attention Dilution ("Lost in the Middle")
                                                                  ├─ Hallucinated Imports & Types
                                                                  └─ Slower & Expensive Edits

Catastrophic Token Exhaustion: Grepping and dumping 8 to 15 raw source files just to discover where a function lives burns 8,000–16,000 tokens per prompt, draining rate limits and budget.
Cognitive Degradation ("Lost in the Middle"): Modern LLMs suffer from severe attention degradation when forced to parse thousands of lines of irrelevant utility and boilerplate code. Reasoning accuracy drops significantly when the context window is flooded with noise.
Hallucinated Contracts: Unable to see concise domain boundaries, the mode