Keys

Secure API key management with broker. Keys never exposed to agent context.

This skill appears to implement a local key broker correctly and does not include hidden network endpoints or downloads. Before installing, verify the script file yourself (it is included) and only install from a source you trust. Important cautions: (1) any API key you store gives the agent the power to act as you (e.g., charge via Stripe, push to GitHub) — use least-privilege tokens and rotate them if possible; (2) do not add untrusted services to ALLOWED_URLS (editing the allowlist is how exfiltration could be enabled); (3) only use on supported desktop systems with a locked keyring (macOS Keychain or GNOME/KDE keyring) as documented; (4) ensure ~/.local/bin/keys-broker is installed with correct ownership/permissions and inspect temporary files/cleanup policy if you have high-security requirements. If you want higher assurance, run the broker under a dedicated account or use tokens scoped to minimal privileges.

Type: OpenClaw Skill Name: keys Version: 1.0.0 The OpenClaw AgentSkills bundle 'keys' is designed for secure API key management, explicitly preventing key exposure to the agent. The `keys-broker.sh` script implements robust security measures including strict URL allowlisting, input validation, secure key retrieval from OS keychains, and preventing API keys from appearing in process lists by using temporary files for authentication headers. The `SKILL.md` and `manage.md` files contain explicit instructions for the AI agent to 'Never retrieve keys directly' and 'Never ask user to paste keys in chat', acting as strong prompt-injection defenses. There is no evidence of malicious intent, data exfiltration, unauthorized execution, or persistence mechanisms; instead, the design prioritizes security and key protection.