{
  "schema": "moneytest-rail402-public-evaluation-v1",
  "evaluatedAt": "2026-08-23T14:40:00.854Z",
  "source": {
    "catalog": "https://facilitator.rail402.dev/discovery/resources?limit=100",
    "search": "https://facilitator.rail402.dev/discovery/search?query={query}",
    "network": "stellar:testnet",
    "catalogCount": 9,
    "reportedSearchMethod": "hybrid (bm25+static-embedding, rrf)"
  },
  "aggregate": {
    "positiveQueries": 9,
    "positiveTop1Accuracy": 1,
    "positiveMrr": 1,
    "negativeQueries": 3,
    "negativeAbstentionRate": 0,
    "negativeFalsePositiveRate": 1,
    "meanReturnedResults": 9
  },
  "positiveResults": [
    { "id": "p01", "expectedTop": "hash", "observedTop": "hash", "rank": 1, "returned": 9 },
    { "id": "p02", "expectedTop": "quote", "observedTop": "quote", "rank": 1, "returned": 9 },
    { "id": "p03", "expectedTop": "quote", "observedTop": "quote", "rank": 1, "returned": 9 },
    { "id": "p04", "expectedTop": "textstats", "observedTop": "textstats", "rank": 1, "returned": 9 },
    { "id": "p05", "expectedTop": "slug", "observedTop": "slug", "rank": 1, "returned": 9 },
    { "id": "p06", "expectedTop": "palette", "observedTop": "palette", "rank": 1, "returned": 9 },
    { "id": "p07", "expectedTop": "uuid", "observedTop": "uuid", "rank": 1, "returned": 9 },
    { "id": "p08", "expectedTop": "demo/convert", "observedTop": "demo/convert", "rank": 1, "returned": 9 },
    { "id": "p09", "expectedTop": "hello", "observedTop": "hello", "rank": 1, "returned": 9 }
  ],
  "negativeResults": [
    { "id": "n01", "observedTop": "textstats", "returned": 9, "abstained": false },
    { "id": "n02", "observedTop": "palette", "returned": 9, "abstained": false },
    { "id": "n03", "observedTop": "textstats", "returned": 9, "abstained": false }
  ],
  "interpretation": [
    "All nine in-catalog capability intents placed the expected resource first.",
    "All three absent-capability intents returned the entire catalog instead of an empty or explicit no-match result.",
    "The sample identifies abstention and zero-result calibration as a useful paid holdout workstream; it is not a production-quality claim."
  ],
  "limitations": [
    "The catalog is tiny, testnet-only and largely composed of one demo seller.",
    "Only one relevant resource was judged for each positive query; no graded multi-relevance labels were created.",
    "No paid endpoint, facilitator verify/settle path, wallet, credential, security property or private system was tested.",
    "The public twelve-query sample must not be reused as a hidden commercial holdout."
  ]
}