Claude Skill

shell-sandbox

Run shell commands with strict limits - prefer workspace sandbox, avoid privileged ops, and quote unsafe output. Use whenever exec is required.

LLM Mart · 0 points · 0 0 0

Virus-scanned Reviewed automatically before listing.

Full trust report

Download navinspire-ia-navin-navin_skills_shell-sandbox-e9c73a3.zip · 0 KB

navinspire-ia/navin 31 4 forks AGPL-3.0 Updated 5d ago

Part of navinspire-ia/navin — 182 skills

Install

skills CLI npx skills add https://github.com/Navinspire-ia/navin/tree/main/navin/skills/shell-sandbox

Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install navinspire-ia-navin@llmmart

Git git clone https://github.com/Navinspire-ia/navin.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole navinspire-ia/navin collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Shell Sandbox

Overview

exec is powerful. Use the smallest command, shortest timeout, and clearest cwd.

Rules of engagement

Prefer built-in tools (read_file, grep, edit) over shell equivalents.
Stay inside the workspace when restriction/sandbox is on.
No sudo, no curling pipes to shells, no rewriting history.
Quote paths; avoid untrusted interpolation.
Cap output - pipe through tail/head when listing huge trees.
Destructive commands require human-approval.

Pattern

intent → exact command → expected success signal → run → interpret

Files (navin)

SKILL.md 849 B

---
name: shell-sandbox
description: Run shell commands with strict limits - prefer workspace sandbox, avoid privileged ops, and quote unsafe output. Use whenever exec is required.
metadata: {"navin":{"emoji":"🧰","category":"devops"}}
---

# Shell Sandbox

## Overview

`exec` is powerful. Use the smallest command, shortest timeout, and clearest cwd.

## Rules of engagement

1. Prefer built-in tools (`read_file`, `grep`, `edit`) over shell equivalents.
2. Stay inside the workspace when restriction/sandbox is on.
3. No `sudo`, no curling pipes to shells, no rewriting history.
4. Quote paths; avoid untrusted interpolation.
5. Cap output - pipe through `tail`/`head` when listing huge trees.
6. Destructive commands require `human-approval`.

## Pattern

```text
intent → exact command → expected success signal → run → interpret
```