Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Add tmux guidance for long-running polling process — prevents SIGKILL from agent exec timeouts
What This Skill Does
Confidant handles secure secret handoff between users and AI agents. Instead of pasting API keys or passwords in chat, the agent generates a one-time web form URL that the user opens in a browser. Secrets are saved to disk and never appear in conversation history.
Secrets never appear in chat logs or shell history because handoff happens through a browser form, not text input.
When to use it
- Setting up an OpenAI API key for a new agent environment
- Passing a database password during dev environment setup
- Storing a GitHub token without exposing it in terminal history
- Collecting credentials from a remote user over a public tunnel
- Onboarding a new service integration without chat-visible secrets
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Can you set up my OpenAI API key?
- 1Run request-secret.sh with --label 'OpenAI API Key', --service openai, and --tunnel flags
- 2Receive the secure one-time URL from script output
- 3Share the URL with the user in chat and wait
- 4Script polls until the user submits the key through the browser form
- 5Confirm the secret is saved to ~/.config/openai/api_key
API key saved to ~/.config/openai/api_key with chmod 600 permissions, never exposed in chat