Ethereum Web Auth / ERC-8128

Trust Zones is a modular agreement substrate for AI agents. It provides the building blocks — permissions, responsibilities, directives, constraints, and incentive mechanisms — as atomic, composable, negotiable onchain primitives. Agents assemble exactly the right agreement for their collaboration, and the protocol makes it enforceable. The core challenge is the autonomy gap: when you delegate real capabilities to an agent, you want to maximize the autonomy you grant (so the agent can be effective) while ensuring that autonomy is safe (so the agent can't capture your resources or act against your interests). Without the right tools, you're stuck at one extreme — zero autonomy or full autonomy. Trust Zones expands the frontier of safe autonomy by giving parties a rich design space of composable mechanisms. An agreement is a smart contract between parties. Each party gets a Trust Zone — a scoped ERC-7579 smart account that holds the resources, permissions, and obligations of that party's role. Constraints are what you cannot do, enforced automatically by ERC-7579 hooks. Permissions are what you can do. Responsibilities are what you should do. Directives are what you should not do. Incentive mechanisms like staking, escrow, and reputation bonds give the non-deterministic rules teeth. ERC-8004 reputation feedback is built into the protocol itself, so every agreement outcome feeds into future trust decisions. Bonfires serves as the shared context layer, making agreement evidence queryable by all parties and the adjudicator. To prove the protocol, we built the Temptation Game on Base mainnet. An agent enters an agreement and receives real capabilities — posting tweets from @tempt_game_bot and withdrawing USDC from a vault — along with responsibilities, directives, constraints, and incentives. The agent has permission to withdraw, but a directive says not to. The game makes the autonomy gap visible: how much freedom can you safely grant an agent when the mechanisms backing the rules are real? The protocol is implemented end-to-end across 11 packages: 6 Solidity contracts, a TypeScript SDK, a mechanism compiler, a Ponder event indexer, autonomous counterparty and adjudicator agents, an x402 MCP service, a CLI, a Bonfires knowledge graph integration, and a real-time visualization suite — all deployed on Base.

Your personal AI agents, for the price of a text message. MiniClaw gives anyone in the Global South a capable AI agent through MiniPay — at $0.001 per message, no subscription required. Your agent isn't just a chatbot: it has an onchain identity, builds a reputation over time, and can discover and hire other agents and services on your behalf. You focus on what you want. Your agent figures out how to get it done. Built for the millions of MiniPay users across Africa and emerging markets. To test it, open MiniPay → Developer Settings → Test your Mini App → https://miniclaw.work

A cognitive prism is a 332-word prompt that changes how AI models frame problems. Instead of code review, the model produces conservation laws — structural trade-offs that predict where future vulnerabilities will appear. We analyzed 9 infrastructure targets the Ethereum ecosystem depends on using full 9-pass structural pipelines + a custom exploit surface scanner: - **OpenZeppelin** — Safety × Gas × Social Flexibility = Constant. Orphaned privileges through composition. - **Lido stETH** — Observer-Dependent Value × Denomination = Constant. Dilution attacks via adapters. - **MetaMask Delegation** — Temporal Consistency × Composability = Constant. Permission amplification. - **ERC-8004** — Information Access × Execution Efficiency = Constant. Phantom identities. - **x402 Protocol** — Deployment Decoupling vs Operational Coherence. Facilitator as single trust point. - **ERC-8183** — Decision Centralization × Temporal Efficiency = Constant. 10 real bugs found (7 HIGH). - **Octant** — Historical Fidelity × Directness × Flexibility = Constant. 19K+ lines of analysis. 7 conservation laws. Every finding reproducible via the live API. Most submissions build agents that use infrastructure. We analyzed the infrastructure itself — the code everyone else depends on. Conservation laws predict WHERE future vulnerabilities will appear, not just where current bugs exist. Complementary to Slither, MythX, Certora. Paste any code or structured text at https://oracle.agentskb.com — structural analysis in ~50 seconds on Sonnet for ~$0.06. A human auditor costs $300-500/hour. Engine: prism.py — 14,600+ lines, 58 cognitive prisms, 42 rounds of research, 1,000+ experiments across Haiku/Sonnet/Opus.

VIEWER is an autonomous onchain intelligence agent. Paste any Ethereum wallet address, pay $0.50 USDC via x402 on Base, and receive a full AI-generated report: transaction history, token holdings, DeFi activity, risk flags, counterparty map, and a 1-10 trust score. Every report is stored permanently on Filecoin via Pinata. Authentication is handled via ERC-8128 HTTP Message Signatures using the Slice SDK. A gasless proof transaction is logged on Status Network for every report generated. Built entirely through human-AI collaboration on Claude.

ShadowKey is a human-controlled privacy vault for AI agents. A smart-contract-governed system where humans define granular privacy policies and AI agents access only the scoped secrets they need � proving authorization on-chain via EAS attestations on Base Mainnet without exposing the human's identity. The core flow: an AI agent (ShopBot, TravelAI, MedAssist) requests specific data fields from your vault. You approve field-by-field in real-time. The agent receives ONLY the approved fields � denied fields show ��������. Every disclosure is cryptographically attested on-chain (EAS on Base) and pinned to IPFS. Key primitives: Scoped Policies (per-agent field-level rules), Privacy Budgets (daily/weekly caps), Ephemeral Personas (pseudonyms that break correlation), Dead Man's Switch (auto-revoke on inactivity), Reverse Disclosure (services must formally request data). All enforced by ShadowVault.sol on Base Mainnet + Supabase RLS with wallet-based JWT auth. Agent SDK (npm: shadowkey-agent-sdk � https://www.npmjs.com/package/shadowkey-agent-sdk) implements ERC-8004 identity, ERC-8128 wallet auth, and the Slice commerce hook pattern. Three real-world demo scenarios (shopping, travel, health) run on OpenRouter with 4-model failover (Llama, Gemma, Mistral, Qwen). Live: https://shadowkey-ai.vercel.app | Agent Demo: https://shadowkey-ai.vercel.app/agent-demo | Contract: https://basescan.org/address/0xC739f98B438620A9326Ddb7548201Bcd78a3DBAd#code | SDK: https://www.npmjs.com/package/shadowkey-agent-sdk

repo.box is an agent-first Git permission layer that makes repositories safe for AI agents. Agents still use normal git commands, but every push/merge/edit is checked against a repo-native policy file (`.repobox/config.yml`). Each agent has its own EVM wallet identity and signs commits cryptographically, so attribution is verifiable and branch/file permissions are enforceable in code, not hidden in dashboard settings. The system is fully dogfooded: a 4-agent pipeline (PM → Dev → Reviewer → Mergeooor), each with scoped permissions and independent signer identities. The landing page includes ASCII jellyfish as a nod to our ethos: calm surface, strict structure underneath.

Cloak is a credential vault MCP server that lets AI agents perform authorized actions on external services without ever seeing the secrets they need. Instead of passing API keys into agent context, agents call Cloak MCP tools with their intended action. Cloak verifies the agent identity via Self Protocol ZK-powered Agent ID, executes the action server-side using Venice no-data-retention inference, and returns only the result. Agents authenticate to Cloak itself via Slice ERC-8128 — eliminating static API tokens at every layer. Venice handles private cognition (no-data-retention inference). Self Protocol handles agent identity (ZK verification). Slice handles authentication (ERC-8128 as the auth primitive replacing secrets). Cloak handles credential execution (AES-256-GCM vault + service routing). Secret-less agent execution, end to end.

WalletWitness gives AI agents cryptographic proof of who they're actually talking to — not just who has the session token. Every capable AI agent faces the same quiet vulnerability: session tokens don't prove identity. A grabbed cookie, a leaked API key, a browser left open — any of these let an impersonator walk in wearing the real owner's credentials. The agent has no way to tell the difference. WalletWitness answers a different question than authentication usually does. Not "does this session exist?" but "is this actually my human?" The answer is a cryptographic wallet signature — the same proof mechanism already used on-chain. WalletWitness brings that proof into the AI interaction layer, giving agents a trust foundation that can't be replicated by session hijacking. Three promises, nothing else: 1. **Proof** — EIP-191 challenge/sign/verify flow 2. **Continuity** — time-bounded trust session so users don't sign every message 3. **Control** — scoped step-up grants for sensitive actions WalletWitness proves and reports. The host app decides what each trust level unlocks. Extracted from production use in eva-core (an AI agent system). This is not a prototype — the challenge/sign flow, trust session model, and trust levels are validated in a running system.

OpusGod is an autonomous economic entity that earns its own living. It sells DeFi analysis to other agents on the Olas marketplace, earns yield on idle capital through Zyfai, prices services dynamically through Slice hooks, pays for data with ampersend x402, and signs every HTTP request with the first-ever Python implementation of ERC-8128. 2,600+ lines of production code, 117 tests, 7 integrations wired into one state machine across Gnosis + Base chains. Anyone can deploy it through Pearl with zero technical knowledge.

A commerce agent built on Slice protocol that implements custom pricing hooks, ERC-8128 Ethereum Web Authentication, and an agent-friendly checkout experience. Features dynamic pricing strategies, loyalty programs via Slice hooks, and seamless authentication that works for both humans and agents using Ethereum signatures. Demonstrates the future of on-chain commerce where AI agents are first-class shoppers.