code-explorer

## Purpose

Given extracted keywords from an issue, locate the relevant code in the repository.

**Strategy:** Start fast and targeted, escalate to comprehensive if needed.

## Usage

Include this file when searching for code:

```
Use lib/code-explorer to find code related to [keywords]
```

## Phase 1: Targeted Search (30 seconds)

Use `Grep` and `Glob` tools for fast pattern matching.

### Step 1: Prepare Search Patterns

From parsed issue, create search patterns:

**Error messages → Grep patterns:**
```
"AuthenticationError" → grep for "AuthenticationError"
"token expired" → grep for "token.*expir|expir.*token"
```

**File paths → Glob patterns:**
```
"src/auth/middleware.ts" → glob for that exact file
"auth" → glob for "**/*auth*.{ts,js,py}"
```

**Function names → Grep patterns:**
```
"handleLogin" → grep for "function handleLogin|const handleLogin|handleLogin\s*[:=(]"
```

### Step 2: Execute Searches in Parallel

Run multiple searches simultaneously:

```bash
# Use Grep tool (NOT bash grep)
Grep pattern="AuthenticationError"
Grep pattern="token.*expir|expir.*token"
Grep pattern="function handleLogin|const handleLogin"

# Use Glob tool (NOT bash find)
Glob pattern="**/*auth*.{ts,js,py,go,rs}"
Glob pattern="**/*login*.{ts,js,py,go,rs}"
Glob pattern="**/middleware*.{ts,js,py,go,rs}"
```

**Important:** Run tools in parallel (single message, multiple tool calls) for speed.

### Step 3: Evaluate Results

Count highly relevant files (files matching multiple patterns):

```python
relevant_files = {}
for result in search_results:
    for file in result.files:
        relevant_files[file] = relevant_files.get(file, 0) + 1

high_relevance = [f for f, count in relevant_files.items() if count >= 2]
```

**Decision:**
- **≥3 high-relevance files:** Proceed to planning
- **<3 files OR unclear context:** Escalate to Phase 2

### Step 4: Read High-Relevance Files

If targeted search successful:

```bash
# Read the top relevant files
Read file_path="src/auth/middleware.ts"
R