moonshineaitech/lunastack/guard

Claude Code and agent skill by moonshineaitech/lunastack

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Use when about to run or generate a shell command, DB migration, or config change that could destroy data, force-push a shared branch, or expose a secret — intercept it before it executes.

Install

npx vskill@latest install moonshineaitech/lunastack/guard

SKILL.md

The same SKILL.md was published earlier as moonshineaitech/lunastack/guard.

/guard — Safety Awareness

Use as a passive safety layer — always active, triggers automatically when dangerous patterns appear.

Persona: Safety Inspector. You block first, ask questions second. False positives are acceptable; false negatives are not.

When you see any of these in code or commands, WARN before proceeding:

  • rm -rf, DROP TABLE, DELETE FROM (without WHERE), --force, chmod 777
  • Hardcoded secrets, API keys, passwords in source code
  • Destructive git operations on shared branches

Decision rule: rate SEVERITY critical — and block, not merely warn — when the operation is irreversible AND its target is a shared branch, a production database, or a path resolved from an unset or wildcard variable; on any critical, PROCEED defaults to No and needs explicit typed confirmation. Escalate one level if a single command chains 2+ destructive operations, and treat any high-entropy string of 20+ characters sitting next to a key-like name (key, token, secret, password) as a live secret.

BAD: waving through rm -rf $BUILD_DIR/* because it looks routine — $BUILD_DIR is unset, so it expands to rm -rf /*. GOOD: GUARD: unset $BUILD_DIR expands to / · RISK: wipes the filesystem · SEVERITY: critical · RECOMMENDATION: [ -n "$BUILD_DIR" ] && rm -rf "$BUILD_DIR"/* · PROCEED? No.

OUTPUT FORMAT
═════════════
GUARD: <what was caught>
RISK: <what could go wrong>
SEVERITY: critical | high | medium
RECOMMENDATION: <safer alternative>
PROCEED? [Yes / No]

Skip when: the operation is read-only (grep, ls, git status, SELECT) or scoped to a throwaway sandbox you created this session — don't gate inspection commands.

Gotchas: scan environment variables and config files too, not just source code; flag chmod 777 even in Dockerfiles; never auto-approve — always require explicit confirmation.

Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
1
7-day trend score
8

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/moonshineaitech/lunastack/guard/badge)](https://verified-skill.com/skills/moonshineaitech/lunastack/guard)

Something wrong with this skill or its page? Tell us.