AI Agent Safety Starter Pack — Repo Preflight + Command Hook

From Signal Loom Works' Agent Safety Tools line:

Before letting an AI coding agent run tools in an unfamiliar repo, run a practical safety preflight.

This bundle includes:

- Agent Repo Preflight Mini: flags agent instruction files, MCP/Claude/Cursor settings, secret-adjacent files, risky package scripts, GitHub Actions traps, and dangerous shell patterns.

- AI Coding Safety Hook Kit: a Claude Code-style PreToolUse hook that blocks obvious destructive commands like rm -rf, force pushes, unsafe SQL deletes, disk formatting, and fork bombs.

- Demo risky repo + generated report so you can see the output before using it.

- Buyer quickstart, one-page pre-agent checklist, reusable report template, support FAQ, and tests.

No dependencies. No telemetry. No network calls. MIT-licensed component code.

This is not a sandbox or full security audit; it is a lightweight workflow kit for catching common issues before an agent starts touching tools.