command-executor

# Command Executor

## Overview

This skill extends the command-verify skill to actually execute safe commands and capture real output. It provides execution capabilities with safety checks and output capture.

## When to Use

Invoke this skill when the user asks to:
- "Actually run the commands"
- "Execute the commands"
- "Test if commands work"
- "Verify commands actually execute"
- "Run the safe commands and show output"

## Safety First

**NEVER auto-execute dangerous commands.** Always follow these safety rules:

### Commands to NEVER Execute:
- `rm -rf` (destructive deletion)
- `git push --force` (destructive git operation)
- `npm run clean/clear/reset` (destructive cleanup)
- `drop database` (database destruction)
- Any command with `--delete` or `--force` flags
- `truncate` operations

### Commands Requiring Confirmation:
- `npm install` (modifies node_modules)
- `npm run format` (modifies source files)
- Commands with `--force` flags
- Any command modifying files

### Commands Safe to Auto-Execute:
- `npm run build`
- `npm run test`
- `npm run lint`
- `npm run typecheck`
- `git status`
- `node --version`
- Read-only operations

## How It Works

### 1. Pre-Flight Checks

Before executing any commands:
- Check git status (ensure working directory is clean)
- Check git branch (warn if on main/master)
- Check for uncommitted changes (ask before running)

### 2. Execution Plan

Always show the user what will be executed:

```
Execution Plan:
✓ npm run build (safe, will execute)
✓ npm run test (safe, will execute)
? npm install (conditional, will ask)
⊝ rm -rf node_modules (dangerous, will skip)

Continue? [y/N]
```

Use the AskUserQuestion tool to get confirmation.

### 3. Command Execution

Execute commands sequentially (not in parallel for safety):
- Capture stdout, stderr, exit code, duration
- Stop on first failure (fail-fast)
- Log all executions to audit log

### 4. Output Capture