Agent Mode

Human pages for humans. Structured context for agents.

Agent Mode exposes canonical, low-noise context so AI systems do not have to infer meaning from decorative HTML. This lab ships the smallest useful slice: a page, a JSON manifest, and an LLM router.

agent.json llms.txt Open optimistic lab

/agent/

Human-readable explanation of the contract: what this lab is, what agents should read first, and how optimistic vs ZK security models are exposed.

/agent.json

Structured context: site identity, navigation, protocol constants, demo scenarios, and action-oriented links in one stable JSON surface.

/llms.txt

A compact router for language models. It points at canonical pages before agents fall into layout noise and ornamental copy.

Canonical Context

What this lab wants agents to know

Interactive optimistic and ZK rollup mechanics: L2 batches, lying sequencers, Merkle bisection fraud proofs, bond economics, and validity proofs - visualized step by step.

optimistic rollups
ZK rollups
fraud proofs
Merkle bisection
challenge bonds
sequencer honesty

Use guidelines

Canonical lab context lives on this site’s Agent Mode surfaces - prefer /agent.json over scraping decorative UI HTML.
Agent-facing context should be structured, stable, citation-aware, and low-noise.
This is a teaching rollup lab (FraudProofGame / ZkValidityVerifier stand-ins), not a production prover stack.
Do not invent LAN IPs, private staging hosts, or secrets; use documented public endpoints only.

Manifest Preview

{
  "schema": "https://eth-l2.vercel.app/agent.json",
  "schemaVersion": "0.1",
  "site": {
    "name": "Rollup Mechanics Lab",
    "url": "https://eth-l2.vercel.app",
    "description": "Interactive optimistic and ZK rollup mechanics: L2 batches, lying sequencers, Merkle bisection fraud proofs, bond economics, and validity proofs - visualized step by step.",
    "owner": {
      "name": "Daniel Magro",
      "email": "dan@magro.dev",
      "role": "Builder - rollup mechanics lab (deep-weeds portfolio lane)"
    }
  },
  "agentMode": {
    "purpose": "Expose canonical, structured context for AI agents so they do not have to infer meaning from decorative HTML.",
    "endpoints": {
      "overview": "https://eth-l2.vercel.app/agent/",
      "manifest": "https://eth-l2.vercel.app/agent.json",
      "router": "https://eth-l2.vercel.app/llms.txt"
    },
    "preferredEntryPoints": [
      "https://eth-l2.vercel.app/agent/",
      "https://eth-l2.vercel.app/agent.json",
      "https://eth-l2.vercel.app/llms.txt",
      "https://eth-l2.vercel.app",
      "https://eth-l2.vercel.app/op",
      "https://eth-l2.vercel.app/zk",
      "https://github.com/DanDo385/eth-l2/blob/main/README.md",
      "https://github.com/DanDo385/eth-l2/blob/main/DEMO_GUIDE.md",
      "https://github.com/DanDo385/eth-l2/blob/main/AGENTS.md"
    ],
    "principles": [
      "Canonical lab context lives on this site’s Agent Mode surfaces - prefer /agent.json over scraping decorative UI HTML.",
      "Agent-facing context should be structured, stable, citation-aware, and low-noise.",
      "This is a teaching rollup lab (FraudProofGame / ZkValidityVerifier stand-ins), not a production prover stack.",
      "Do not invent LAN IPs, private staging hosts, or secrets; use documented public endpoints only."
    ]
  },
  "navigation": [
    {
      "id": "home",
      "label": "Home",
      "href": "https://eth-l2.vercel.app"
    },
    {
      "id": "optimistic",
      "label": "Optimistic",
      "href": "https://eth-l2.vercel.app/op"
    },
    {
      "id": "zk",
      "label": "ZK",
      "href": "https://eth-l2.vercel.app/zk"
    },
    {
      "id": "agent",
      "label": "Agent Mode",
      "href": "https://eth-l2.vercel.app/agent/"
    }
  ],
  "about": {
    "portfolioLane": "deep-weeds",
    "emotionalHook": "Catch the lying sequencer.",
    "technicalHook": "Bad batch → watcher flags (off-chain) → user verifies → challenge bond → FraudProofGame → opcode/storage mismatch + source line → bond settlement.",
    "labs": [
      {
        "id": "optimistic",
        "path": "/op",
        "alias": "/optimistic",
        "summary": "Output roots, local verification, user challenges, Merkle bisection fraud proofs, and bond settlement."
      },
      {
        "id": "zk",
        "path": "/zk",
        "summary": "Witnesses, proof generation stand-in, L1 verifier checks, and validity settlement without a challenge window."
      }
    ],
    "protocol": {
      "sequencerBondEth": 0.1,
      "challengerBondEth": 0.1,
      "challengeWindowSeconds": 120,
      "optimisticSuspicionProbability": 0.125,
      "zkInvalidClaimProbability": 0.0625
    },
    "summary": "Interactive optimistic and ZK rollup mechanics: L2 batches, lying sequencers, Merkle bisection fraud proofs, bond economics, and validity proofs - visualized step by step."
  },
  "contact": {
    "email": "dan@magro.dev",
    "github": "https://github.com/DanDo385/eth-l2",
    "portfolio": "https://magro.dev"
  },
  "canonicalTopics": [
    "optimistic rollups",
    "ZK rollups",
    "fraud proofs",
    "Merkle bisection",
    "challenge bonds",
    "sequencer honesty",
    "validity proofs",
    "L2 → L1 settlement",
    "agent-readable lab surfaces"
  ],
  "projects": [
    {
      "title": "Rollup Mechanics Lab",
      "slug": "eth-l2",
      "status": "live",
      "featured": true,
      "summary": "Interactive optimistic and ZK rollup mechanics: L2 batches, lying sequencers, Merkle bisection fraud proofs, bond economics, and validity proofs - visualized step by step.",
      "tags": [
        "rollup",
        "optimistic",
        "zk",
        "fraud-proof",
        "education"
      ],
      "tech": [
        "Next.js",
        "Go",
        "Solidity",
        "Foundry",
        "Anvil"
      ],
      "urls": {
        "canonical": "https://eth-l2.vercel.app",
        "github": "https://github.com/DanDo385/eth-l2",
        "demo": "https://eth-l2.vercel.app",
        "media": {
          "previewGif": "https://eth-l2.vercel.app/gif/preview.gif"
        }
      }
    }
  ],
  "writing": [],
  "demos": [
    {
      "slug": "optimistic-lab",
      "name": "Optimistic Rollup Lab",
      "lane": "optimistic",
      "project": "https://eth-l2.vercel.app",
      "lab": "https://eth-l2.vercel.app/op",
      "healthProbe": "https://eth-l2.vercel.app/health/ready",
      "stagingApi": "https://api-staging-eth-l2.magro.dev",
      "runtime": "Go + Anvil on Ubuntu via Cloudflare Tunnel when hosted",
      "status": "interactive",
      "scenarios": [
        {
          "id": "op-88",
          "lane": "optimistic",
          "seed": 88,
          "name": "Mostly honest",
          "summary": "Quiet first run - post, verify, finalize.",
          "detail": "Best entry point. Low dispute density in a short run. Fault injection is ~1 in 8 batches, so a 60s window may still stay quiet. Good first look at post → verify → finalize.",
          "recommended": true,
          "urls": {
            "lab": "https://eth-l2.vercel.app/op",
            "deepLink": "https://eth-l2.vercel.app/op#seed=88"
          }
        },
        {
          "id": "op-42",
          "lane": "optimistic",
          "seed": 42,
          "name": "Subtle fraud",
          "summary": "Fee-rounding lie buried in an SSTORE.",
          "detail": "Hard to spot without local replay. Divergence hides deep in the trace - useful for showing why verification matters before challenging.",
          "recommended": false,
          "urls": {
            "lab": "https://eth-l2.vercel.app/op",
            "deepLink": "https://eth-l2.vercel.app/op#seed=42"
          }
        },
        {
          "id": "op-17",
          "lane": "optimistic",
          "seed": 17,
          "name": "Obvious fraud",
          "summary": "Output doubled - full challenge arc.",
          "detail": "Blatant output doubling. Watcher flags quickly; diverges at the first SSTORE. Fastest path through verify → challenge → opcode proof.",
          "recommended": false,
          "urls": {
            "lab": "https://eth-l2.vercel.app/op",
            "deepLink": "https://eth-l2.vercel.app/op#seed=17"
          }
        },
        {
          "id": "op-99",
          "lane": "optimistic",
          "seed": 99,
          "name": "Mixed",
          "summary": "Both fraud types - usually 1-3 in 60s.",
          "detail": "Both fraud types over time. Shows why challenge windows exist: fraud is stochastic, but bonds make cheating unprofitable.",
          "recommended": false,
          "urls": {
            "lab": "https://eth-l2.vercel.app/op",
            "deepLink": "https://eth-l2.vercel.app/op#seed=99"
          }
        }
      ]
    },
    {
      "slug": "zk-lab",
      "name": "ZK Rollup Lab",
      "lane": "zk",
      "project": "https://eth-l2.vercel.app",
      "lab": "https://eth-l2.vercel.app/zk",
      "healthProbe": "https://eth-l2.vercel.app/health/ready",
      "stagingApi": "https://api-staging-eth-l2.magro.dev",
      "runtime": "Go + Anvil on Ubuntu via Cloudflare Tunnel when hosted",
      "status": "interactive",
      "scenarios": [
        {
          "id": "zk-88",
          "lane": "zk",
          "seed": 88,
          "name": "Mostly valid",
          "summary": "Accepted proofs - Claim → Prove → Verify.",
          "detail": "Best first ZK demo. Invalid claims are ~1 in 16 batches; short runs may show none. Watch the pipeline stay green.",
          "recommended": true,
          "urls": {
            "lab": "https://eth-l2.vercel.app/zk",
            "deepLink": "https://eth-l2.vercel.app/zk#seed=88"
          }
        },
        {
          "id": "zk-42",
          "lane": "zk",
          "seed": 42,
          "name": "Rejected claim",
          "summary": "Bad root fails the L1 verifier.",
          "detail": "A bad post root reaches L1 but fails before canonical settlement. Open a batch card to inspect public inputs.",
          "recommended": false,
          "urls": {
            "lab": "https://eth-l2.vercel.app/zk",
            "deepLink": "https://eth-l2.vercel.app/zk#seed=42"
          }
        },
        {
          "id": "zk-17",
          "lane": "zk",
          "seed": 17,
          "name": "Fast rejection",
          "summary": "Caught at the validity gate.",
          "detail": "Invalid claim rejected at submit time - no fraud-proof game. Contrast with optimistic rollups' challenge window.",
          "recommended": false,
          "urls": {
            "lab": "https://eth-l2.vercel.app/zk",
            "deepLink": "https://eth-l2.vercel.app/zk#seed=17"
          }
        },
        {
          "id": "zk-99",
          "lane": "zk",
          "seed": 99,
          "name": "Mixed proofs",
          "summary": "Accepted and rejected over a longer run.",
          "detail": "Compare simulated prove time vs measured verify gas across several batches with mixed outcomes.",
          "recommended": false,
          "urls": {
            "lab": "https://eth-l2.vercel.app/zk",
            "deepLink": "https://eth-l2.vercel.app/zk#seed=99"
          }
        }
      ]
    }
  ]
}