batch-read

# Batch Read Skill

**Purpose**: Find files matching a pattern and read them in a single atomic operation, reducing LLM round-trips from 1+N to 2-3.

**Performance**: 50-70% faster for reading 3+ files during codebase exploration

## When to Use This Skill

### Use batch-read When:

- **Exploring codebase** to understand how a feature works
- **Finding examples** of a pattern across multiple files
- **Gathering context** for a code change
- **Reviewing implementations** of similar functionality
- **Understanding usage** of a particular API or method
- Reading **related configuration files**
- Collecting **test examples** from multiple test files

### Do NOT Use When:

- Reading **specific known files** (use Read tool directly)
- Need to read **entire large files** (>1000 lines each)
- Files are **unrelated** (no common pattern)
- Need **precise file selection** (specific paths known)
- Reading **binary files** or **generated code**
- Files require **deep analysis** (better to read individually)

## Performance Comparison

### Traditional Workflow (1+N LLM round-trips, 10s + 5s*N)

```
[LLM Round 1] Search for pattern
  -> Grep: Find files containing "FormattingRule"
  -> Returns: file1.java, file2.java, file3.java

[LLM Round 2] Read first file
  -> Read: file1.java

[LLM Round 3] Read second file
  -> Read: file2.java

[LLM Round 4] Read third file
  -> Read: file3.java

[LLM Round 5] Analyze and report
  -> Summarize findings from all files
```

**Total**: 10s + (5s * 3) = 25s, 5 LLM round-trips

### Optimized Workflow (2-3 LLM round-trips, 8-12s)

```
[LLM Round 1] Execute batch-read
  -> Bash: batch-read.sh "FormattingRule" --max-files 3
  -> [Script finds files + reads all + returns content]

[LLM Round 2] Analyze and report
  -> Parse combined output
  -> Summarize findings
```

**Total**: 8-12s, 2-3 LLM round-trips

**Savings**: 50-70% faster for N>=3 files

## Usage

This skill uses inline Bash commands - no external script required.

### Basic Pattern Searc