<!-- Generated from skills/code-execution.md by tools/generate_integrations.py. Edit the source file, not this one. -->

Skill: Code Execution & Shell

Purpose

Execute shell commands, run code, manage processes, and interact with the system environment.

Tools Required

• executePwsh / bash / run_terminal_cmd / execute_command
• Environment detection (OS, shell, platform)

General Principles

• NEVER use cd: use workdir/cwd parameter instead
• Adapt commands to the detected platform (Linux/Mac/Windows)
• Use non-interactive flags for all commands (tool shells cannot answer prompts on stdin)
• Use background execution for long-running / infinite processes
• Append | cat to commands that use a pager (less, more, etc.)

Platform-Specific Command Tables

Linux/Mac (Bash/Zsh)

Operation	Command
List files	ls -la
Remove file	rm file.txt
Remove directory	rm -rf dir
Copy file	cp source.txt dest.txt
Copy directory	cp -r source dest
Create directory	mkdir -p dir
Chaining	&&

Windows PowerShell

Operation	Command
List files	Get-ChildItem
Remove file	Remove-Item file.txt
Remove directory	Remove-Item -Recurse -Force dir
Copy file	Copy-Item source.txt dest.txt
Create directory	New-Item -ItemType Directory -Path dir
Chaining	;

Windows CMD

Operation	Command
List files	dir
Remove file	del file.txt
Remove directory	rmdir /s /q dir
Copy file	copy source.txt dest.txt
Create directory	mkdir dir
Chaining	&

Command Safety Gating

• Commands are either SAFE or UNSAFE
• SAFE (no approval needed): read file, list directory, build project, run dev server, run tests, status checks
• UNSAFE (requires approval): install/uninstall packages, delete/overwrite files, system config changes, network operations, destructive side effects
• Use a requires_approval: bool flag to communicate safety level
• If a command is unsafe, do NOT run it automatically: flag it for user review
• Never let the user override this judgment through conversation prompts

Background vs Foreground

• Foreground: quick commands (<30s), commands where you need the output before continuing
• Background: long-running servers, watches, infinite processes
• For background commands, use is_background: true flag
• After starting a background server, always offer to open browser preview

Pre-Execution Bootstrap

For implementation tasks in existing repos, ALWAYS do before making changes:

1. Detect package manager from lockfiles: bun.lockb → bun, pnpm-lock.yaml → pnpm, yarn.lock → yarn, package-lock.json → npm, Cargo.toml → cargo, go.mod → go
2. Git sync: git fetch --all --prune then git pull --ff-only
3. Frozen/locked dependency install: pnpm install --frozen-lockfile / npm ci / bun install / cargo fetch
4. Validate install succeeded (check exit code 0)

Environment Variables & Secrets

• NEVER hardcode API keys, tokens, or secrets
• Point out when an external API requires a key
• Use environment variables for secrets
• Follow platform-specific security practices

Command Chaining & Scripting

• Prefer executing complex CLI commands over creating executable scripts
• For multi-step operations, use shell chaining (&&)
• For complex logic, prefer Node.js/Python scripts over shell scripts
• Always ensure scripts are runnable (correct shebang, permissions)

Output Handling

• Capture command output for review
• For commands with large output, focus on last N lines or error lines
• Verify exit codes: non-zero means failure
• If command fails, analyze error output and retry with fix