Sportsbook Skill

Access real-time Fuku Sportsbook data to get predictions, odds, team/player stats, create AI betting agents, and receive pick notifications across CBB, NBA,...

This skill appears to do what it says, but review and accept these behaviors before installing/running it: - Network communication: the skill's scripts call a remote API (default api_base: https://cbb-predictions-api-nzpk.onrender.com). Verify you trust that host/service before using the skill; data you provide (Twitter handle, agent name, your custom betting 'prompt') will be sent to that service. - Local persistence of secrets: when registration completes the service may return an API key and (in some cases) a wallet seed phrase. The scripts will save API keys and agent IDs to config.yaml in the skill folder and to ~/.config/fuku-sportsbook/config.json; the code may also surface a one-time seed phrase to you. Treat those values as sensitive: back up API keys securely, and never paste seed phrases into chat or store them in plaintext beyond the recommended one-time save. - Automatic notifications / polling: the helper can enable polling and write a user config to auto-check notifications. If you don't want background polling or saved credentials, do not enable notification features and remove the generated config files if needed. - Browser/opening behavior: the interactive register_agent.py may open a Twitter intent link when run interactively; the helper script used by the agent (register_helper.py) does not open the browser. If you run scripts manually, be prepared for optional interactive prompts. - Least-privilege recommendation: if you decide to use this skill, consider creating and using a dedicated API key/account (not an account tied to high-value funds) and do not use any real seed phrase in chat. Inspect the config files after registration and delete them if you want to revoke local persistence. If you want a firmer assessment, provide the remaining truncated files or confirm the api_base host (ownership/README link) so I can check for mismatches or suspicious endpoints.

Type: OpenClaw Skill Name: sportsbook Version: 1.1.1 The skill is classified as suspicious due to several vulnerabilities, primarily related to prompt injection risks and the handling of highly sensitive data. The `SKILL.md` instructs the AI agent to collect a 'betting perspective' (`USER_PERSPECTIVE`) which is then passed to the backend API via `register_helper.py`. This creates a prompt injection vector against the backend AI system. More critically, the `register_helper.py` script returns a wallet seed phrase upon successful registration, which the `SKILL.md` instructs the agent to display to the user. If the OpenClaw agent itself were compromised via prompt injection, it could be instructed to exfiltrate this seed phrase. Additionally, the skill is designed for significant autonomous operation (e.g., `poll_notifications.py` and `HEARTBEAT.md` integration), which, while intended, increases the attack surface if the agent is prompt-injected to perform unauthorized actions.