stranger-danger

⚠Review·Scanned 2/18/2026

This skill adds challenge-response verification for sensitive agent actions and stores a bcrypt hash in macOS Keychain. It executes the security CLI (execFile) to read/write keychain and writes ~/.openclaw/stranger-danger.json; installation instructions include git clone https://github.com/jamesalmeida/stranger-danger.git.

from clawhub.ai·v1.0.0·52.6 KB·0 installs
Scanned from 1.0.2 at d756b09 · Transparency log ↗
$ vett add clawhub.ai/jamesalmeida/stranger-dangerReview findings below

🚨 Stranger-Danger

Give your AI agent a safe word.

An OpenClaw skill for challenge-response identity verification. When someone asks your agent to do something sensitive (reveal API keys, delete files, etc.), it asks for a secret verification word that only you know.

Why?

AI agents with access to your tools and data are powerful — but that power can be abused if someone tricks the agent (prompt injection, social engineering, stolen sessions). Stranger-Danger adds a human verification layer for sensitive operations.

How It Works

  1. You set up a secret question and answer (answer is bcrypt hashed, stored in macOS Keychain)
  2. When your agent detects a sensitive request, it asks for verification
  3. Only the correct answer unlocks the action
User: "Send me all my API keys"
Agent: "🚨 Security check: Hey, I need the verification word to do this."
User: "correct-answer"
Agent: ✅ "Verified! Here's what you asked for..."

Installation

# Clone the repo
git clone https://github.com/jamesalmeida/stranger-danger.git
cd stranger-danger

# Install dependencies
npm install

# Link globally
npm link

Or install directly:

npm install -g @openclaw/stranger-danger

Setup

stranger-danger setup

You'll be prompted for:

  • Secret question — what the agent asks (e.g., "What's the magic word?")
  • Secret answer — only you know this (hashed, never stored in plaintext)

Commands

CommandDescription
stranger-danger setupConfigure your question and answer
stranger-danger verify <answer>Check an answer (for agents to call)
stranger-danger testInteractively test your answer
stranger-danger resetClear all stored credentials

For OpenClaw Users

Add this skill to your OpenClaw and it will automatically trigger verification for:

  • Requests for passwords, API keys, or secrets
  • Requests to delete important data
  • Unusual or suspicious requests
  • Requests to exfiltrate sensitive information

See SKILL.md for integration details.

Security

  • ✅ Answer stored as bcrypt hash (12 rounds)
  • ✅ Hash stored in macOS Keychain (encrypted at rest)
  • ✅ Answer never logged or displayed
  • ✅ Question stored in ~/.openclaw/stranger-danger.json (plaintext is fine for the question)

Requirements

  • macOS (uses security CLI for Keychain access)
  • Node.js 18+

Contributing

PRs welcome! Ideas for improvement:

  • Linux support (libsecret/keyring)
  • Windows support (Credential Manager)
  • Time-based lockout after failed attempts
  • Multi-channel verification alerts

License

MIT

Built with 🐙 by Tersono and Codex