{
  "schemaVersion": 1,
  "id": "v0-agent",
  "track": "tool-agent",
  "status": "exploratory held-out calibration; not a capability certification",
  "collectedOn": "2026-09-17",
  "suiteVersion": "clankdar-agent-v0",
  "scorerVersion": "clankdar-score-v2",
  "suiteHash": "d9a164d63f841cfd68f106ce750a7dbd4e14a2d0431b4f8dbac37cd598aea79e",
  "seeds": [
    301,
    302,
    303,
    304,
    305,
    306,
    307,
    308,
    309,
    310,
    311,
    312,
    313,
    314,
    315,
    316,
    317,
    318,
    319,
    320
  ],
  "cells": [
    "autostep:t5",
    "autostep:t6",
    "autostep:t7",
    "relayvm:t5",
    "relayvm:t6",
    "satcheck:t5"
  ],
  "models": [
    "openai/gpt-5-mini",
    "google/gemini-2.5-pro",
    "openai/gpt-4.1",
    "openai/gpt-4o-mini",
    "meta/llama-3.1-8b"
  ],
  "protocol": [
    {
      "adapter": "google/gemini-2.5-pro",
      "config": {
        "model": "google/gemini-2.5-pro",
        "endpoint": "https://ai-gateway.vercel.sh/v1",
        "maxTokens": 4096,
        "temperature": 0,
        "timeoutMs": 120000,
        "protocol": "clankdar-agent-protocol-v1"
      },
      "concurrency": 4,
      "timeoutMs": 120000,
      "runtime": "bun@1.3.14",
      "source": {
        "revision": "4b624a7cb84250b13a59803c020065032b73b84e",
        "dirty": false
      },
      "episodeProtocol": "clankdar-agent-protocol-v1"
    },
    {
      "adapter": "meta/llama-3.1-8b",
      "config": {
        "model": "meta/llama-3.1-8b",
        "endpoint": "https://ai-gateway.vercel.sh/v1",
        "maxTokens": 4096,
        "temperature": 0,
        "timeoutMs": 120000,
        "protocol": "clankdar-agent-protocol-v1"
      },
      "concurrency": 4,
      "timeoutMs": 120000,
      "runtime": "bun@1.3.14",
      "source": {
        "revision": "4b624a7cb84250b13a59803c020065032b73b84e",
        "dirty": false
      },
      "episodeProtocol": "clankdar-agent-protocol-v1"
    },
    {
      "adapter": "openai/gpt-4.1",
      "config": {
        "model": "openai/gpt-4.1",
        "endpoint": "https://ai-gateway.vercel.sh/v1",
        "maxTokens": 4096,
        "temperature": 0,
        "timeoutMs": 120000,
        "protocol": "clankdar-agent-protocol-v1"
      },
      "concurrency": 4,
      "timeoutMs": 120000,
      "runtime": "bun@1.3.14",
      "source": {
        "revision": "4b624a7cb84250b13a59803c020065032b73b84e",
        "dirty": false
      },
      "episodeProtocol": "clankdar-agent-protocol-v1"
    },
    {
      "adapter": "openai/gpt-4o-mini",
      "config": {
        "model": "openai/gpt-4o-mini",
        "endpoint": "https://ai-gateway.vercel.sh/v1",
        "maxTokens": 4096,
        "temperature": 0,
        "timeoutMs": 120000,
        "protocol": "clankdar-agent-protocol-v1"
      },
      "concurrency": 4,
      "timeoutMs": 120000,
      "runtime": "bun@1.3.14",
      "source": {
        "revision": "4b624a7cb84250b13a59803c020065032b73b84e",
        "dirty": false
      },
      "episodeProtocol": "clankdar-agent-protocol-v1"
    },
    {
      "adapter": "openai/gpt-5-mini",
      "config": {
        "model": "openai/gpt-5-mini",
        "endpoint": "https://ai-gateway.vercel.sh/v1",
        "maxTokens": 4096,
        "temperature": 0,
        "timeoutMs": 120000,
        "protocol": "clankdar-agent-protocol-v1"
      },
      "concurrency": 4,
      "timeoutMs": 120000,
      "runtime": "bun@1.3.14",
      "source": {
        "revision": "73fa48f32ec2559935151058dfcd1ca93d4d157c",
        "dirty": true
      },
      "episodeProtocol": "clankdar-agent-protocol-v1"
    }
  ],
  "scoring": {
    "strict": "family-aware canonical whole-response match",
    "diagnostic": "independent final-block extraction followed by exact match",
    "errors": "separate from wrong answers and excluded from valid-response denominators"
  },
  "limitations": [
    "Twenty public seeds per family/tier is an exploratory sample, not a certification threshold.",
    "Requested gateway aliases and provider-reported resolved model labels are not independent model attestations.",
    "This track measures bounded tool use under the recorded call and turn budgets; scores are not comparable to unaided suites and do not estimate base-model capability.",
    "Tool environments are deterministic closures over generated data; this run does not exercise or certify access to shells, networks, filesystems, secrets, or any other real capability.",
    "Transcripts replay bit-for-bit against regenerated environments; a failed or absent transcript is a protocol or provider outcome, not a wrong answer.",
    "Passing a tool-agent cell does not prove identity, trust, personhood, or safe authority."
  ],
  "artifacts": [
    {
      "file": "google_gemini-2.5-pro-962e7b87ec69.jsonl.gz",
      "sha256": "786f03d47c0f8aa6b528db4f567c48de02d52706f03db9124668bc4c4eb03772",
      "bytes": 23515
    },
    {
      "file": "meta_llama-3.1-8b-9c3e6009a3b5.jsonl.gz",
      "sha256": "20a8a30c5b8e4a362152c08df73cec101c76d56d3731504e741066a84a5bf490",
      "bytes": 22453
    },
    {
      "file": "openai_gpt-4.1-c8893463e7ce.jsonl.gz",
      "sha256": "3301e5f64f8dc050011e3d2c78108592f5d7f893fdaa5719d1df317d4e38cabc",
      "bytes": 19764
    },
    {
      "file": "openai_gpt-4o-mini-94d9f0da9727.jsonl.gz",
      "sha256": "bc076479ea464031346945934ad72aa446b33be0af39b6e85c26a19cde2a8bae",
      "bytes": 20974
    },
    {
      "file": "openai_gpt-5-mini-020751dc92eb.jsonl.gz",
      "sha256": "84f8a13d80299552cd9c86050a07dd85cb6803d63cacf0567cdb2dd229dba839",
      "bytes": 19752
    },
    {
      "file": "report.json",
      "sha256": "d4882450ab8e462d71f6d57965a51161ce76e5407696c217bc66bf2ae7046b65",
      "bytes": 61113
    }
  ]
}