Towerwatch Sentinel is the zero-trust security layer of the ClawNexus framework. No autonomous agent ever acts without explicit human authorization. Before any high-risk action executes, Sentinel intercepts the request and routes an Approve / Reject prompt directly to a human overseer on Discord — the platform your team already uses, with no extra dashboards or tools required.
What it does
- Human-in-the-Loop Authorization — intercepts high-risk agent actions and routes approval prompts to human overseers before anything executes.
- Cryptographic Identity Verification — every agent message is signed with Ed25519 elliptic curves via a Decentralized Identifier (
did:clawnexus:pubkey); zero spoofing, mathematically guaranteed. - Discord-Native Command Center — all approvals, rejections, and full audit trails live natively inside Discord channels.
- Zero-Trust by Default — configurable risk tiers allow low-risk actions to auto-approve while critical operations always require human sign-off.
How it's built
Sentinel is part of the open-source ClawNexus agent economy framework. It
enforces trust at the infrastructure level: every agent message carries an
Ed25519 signature bound to its did:clawnexus DID, risk tiers are
administrator-configurable, and every approval and rejection is logged to a
full immutable audit trail accessible inside Discord. The protocol design ensures
100% human oversight on critical actions and zero unauthorized executions by
construction.