Codebase Analysis

Perform deep codebase exploration and analysis in an isolated sub-agent context. Heavy Grep/Glob operations stay separate from your main work.

Built-in Analysis Types

Type What it finds

dependencies

Import/require graph, circular dependencies, unused imports

dead-code

Exported but never-imported functions, unreachable code paths

api-surface

Public exports, REST endpoints, RPC methods

type-safety

any types, type assertions, missing return types

error-handling

Uncaught promises, empty catch blocks, missing error boundaries

test-coverage

Files without corresponding test files, untested exports

security

Hardcoded secrets, unsanitized inputs, eval usage

patterns

Custom pattern matching (permissions, money, DTOs, etc.)

Usage

/codebase-analysis --type=<type> [--output=summary|detailed]

Examples:

/codebase-analysis --type=type-safety # Find all any types and unsafe casts /codebase-analysis --type=dead-code # Find unused exports /codebase-analysis --type=error-handling # Audit error handling patterns /codebase-analysis --type=patterns # Custom pattern (prompted interactively)

Tools Available in Fork Context

Tool Purpose

tools mcp-tsc <file>

TypeScript diagnostics per file via persistent LSP

tools mcp-ripgrep

Code search MCP server for structured queries

tools collect-files-for-ai <dir> -c N

Gather top N relevant files for analysis

tools files-to-prompt <dir> --cxml

Generate structured context XML

How It Works

• Launches isolated agent -- intensive searching runs in parallel

• Performs extensive Grep/Glob without blocking main session

• Analyzes patterns and cross-references findings

• Returns structured report to main session

• You continue working while analysis runs

Report Format

[Analysis Type] Report

Scanned: N files | Findings: N issues | Severity: High/Medium/Low

Finding 1: [description]

• File: src/utils/format.ts:L45
• Issue: [what's wrong]
• Suggestion: [how to fix]

Summary