E Rich

The swarm-contract skill for OpenClaw is live for testing and is available for manual download and experimentation on swarmcontract.com We are providing that under the MIT license. It is aimed to be a start for the legal firewall or layer in AI agentic transactions. But it is imperfect. It awaits machine to machine agreements for which we also provide a builder on the site for experimentation. I’ve spent the last 30+ years as a technology lawyer — starting before the commercial internet existed. I’ve litigated cases involving shrink-wrap licenses, cloud storage, digital copyright, privacy, and consumer protection against some of the largest companies in the world. I was involved in building out legal compliance for the blockchain. Throughout all of it, the same pattern repeats: new technology outpaces the legal infrastructure designed to govern it, and someone has to start building the bridge. We’re watching it happen again right now with autonomous AI agents. AI agents are browsing the web, signing up for services, and completing purchases on behalf of humans. Every one of those transactions involves legal agreements. None of them are being read. The agent clicks “I Agree” and moves on. The human it represents has no idea what they just agreed to — and under current law, they’re bound by it anyway. The short version: the legal frameworks governing electronic consent were written in 1999 for shopping carts, not for probabilistic AI systems making autonomous legal decisions at scale. So we vibe coded an initial response. Several, actually. Here’s what’s on swarmcontract.com today: 1. The swarm-contract Skill — an open-source legal firewall for OpenClaw that aims to give AI agents the ability to scan, analyze, and evaluate a vendor’s terms and privacy policies before transacting. Human-in-the-loop. Tries to get human authorization before agent accepts. It attempts to save every online agreement and related data the AI agent ultimately “agreed to.” The skill is MIT licensed. A note of caution: each scan involves significant compute and API calls and the results are inconsistent, so we are publishing it for manual installation, experimentation, and testing only at this stage — not for mass or high-volume use. 2. The Swarm Contract Builder — an experimental app for vendors to generate machine-readable Swarm Contracts for their websites, so AI agents can discover and cross-reference terms automatically. The hash aspect of a machine readable agreement is important to reduce redundancy in compute and api calls. 3. The full article explaining the architecture, the reasoning, the limitations, and the vision — including why we called it “swarm” contract and how reliable hash-based machine to machine contracts could make the ecosystem faster as it scales. All of it is experimental. All of it is provided as-is, not legal advice. The AI analysis is probabilistic and can produce varying results. Humans — and their human lawyers — must stay in the loop and do their own review. But some data is better than no data. Some review is better than no review. And an imperfect signal that prompts a human to pay attention to AI agent legal transactions is better than the current default of clicking through blindly. This is a start. It will get better with community input. swarmcontract.com #AI #AgenticAI #SwarmContract #LegalTech #OpenSource #AIAgents #LawAndTechnology #InformedConsent #VibeCoding #OpenClaw @steipete