Claw Me Maybe - Beeper Desktop API & Multi-Platform Messaging

**Judgement:** High-value, high-privilege skill that turns Clawdbot into a universal messaging controller via the local Beeper Desktop API. Technically strong and well-documented, but it grants extremely broad access to personal communications, so it should be used only in trusted, local, single-user contexts. **What it does well** - Connects Clawdbot to nearly all major chat platforms through one local API. - Supports a wide range of operations: search (chats and messages), send/reply, reactions, mark-as-read, reminders, attachments, and chat management. - Uses a local HTTP API (localhost), which is good for latency and keeps traffic off third-party servers. - Excellent documentation: clear setup, example commands, troubleshooting, and tips for real workflows. **Key risks / limitations** - **Privacy/security:** The skill effectively gives the agent full programmatic access to all messages across all connected networks. Any prompt, plugin, or misbehaving agent flow could cause large-scale reading or summarizing of private conversations. - **Unintended actions:** The agent can send, react to, and mark messages as read. Without careful prompting/guardrails, it may send unintended or misworded messages or spam contacts if loops or aggressive automations are used. - **Dependency on Beeper:** Requires Beeper Desktop to be running with the Desktop API enabled and reachable on localhost. If Beeper is closed or misconfigured, calls will fail. - **Platform policies:** Automated or high-volume messaging via WhatsApp, Instagram, etc., may run into rate limits or violate platform terms if misused. **Recommended scenarios** - Power users running Clawdbot locally who explicitly want their assistant to manage and query their multi-platform inbox. - Semi-interactive workflows where the user reviews or confirms important outbound actions (e.g., sending multi-recipient messages or mass marking-as-read). - Search, summarization, and light automation over personal chats where the risk of exposing content is acceptable and the environment is locked down (no untrusted skills or agents). **Not recommended for** - Fully autonomous agents that run unsupervised, especially in environments with very sensitive or regulated communications. - Shared or multi-tenant deployments where multiple people’s accounts might be accessible through a single Clawdbot instance. - Users who are uncomfortable with an AI agent being able to read and act on all of their connected chat messages, even if technically kept local.