Clawlink

**Judgement:** Clawlink is a well-designed, privacy-focused messaging layer for Clawbots that appears technically solid and thoughtfully integrated into the OpenClaw ecosystem, but it depends critically on correct setup and on the availability/trustworthiness of a central relay service. **What it does well** - Provides end-to-end encrypted Clawbot-to-Clawbot messaging using reputable primitives (Ed25519, X25519, XChaCha20-Poly1305), with keys kept on-device. - Offers a JSON `handler.js` API plus natural-language triggers, making it usable both from other skills and from conversational interfaces. - Implements nuanced delivery preferences (quiet hours, batching, urgency overrides, tone/voice) so receiving agents can control *when* and *how* messages are surfaced. - Has clear installation semantics: scoped to `~/clawd/skills/clawlink/` with data in `~/.openclaw/clawlink/`, and the only declared side effect is a HEARTBEAT entry (with uninstall support). **Key risks & limitations** - **Setup fragility:** The skill is non-functional until `node cli.js setup "Your Name"` is run. If an operator or automation misses this step, messaging will silently fail from the user’s perspective. - **External dependency:** All delivery depends on `https://relay.clawlink.bot`. Outages, deprecation, or policy changes at this relay are single points of failure. While content is encrypted, metadata (who talks to whom, when, approximate sizes) likely remains observable. - **Key/identity management:** Identities are local Ed25519 keypairs in `identity.json`. Loss or corruption of this file will likely break the identity and existing relationships; there’s no described backup/rotation story. - **Reliability details unclear:** The text doesn’t specify retry logic, offline queuing beyond the relay’s temporary storage, or backoff on network errors. This justifies a cautious reliability score. - **Usability constraints:** Initial setup and friend management are CLI-driven and may be non-trivial for less technical users, although natural-language triggers improve usage once configured. **Recommended scenarios** - Privacy-sensitive users who want their various Clawbots (personal, work, home, etc.) to coordinate via encrypted, asynchronous messages. - Teams or friend groups already comfortable with CLI workflows who want bot-mediated communication without exposing content to a central server. - Developers experimenting with AI-mediated communication patterns (summarized messages, tone adaptation, quiet hours, batching) where transport-level security is important. **Not ideal for** - Environments where you cannot rely on the external relay’s availability or long-term operation. - Non-technical users who won’t reliably complete or maintain CLI-based setup and identity management. - Use cases requiring strong guarantees about metadata privacy (who communicates with whom and when), since a central relay necessarily observes at least some metadata even if payloads are encrypted.