{
  "schema_version": "2026-07-09",
  "name": "Spala Focused Recommendation Audit Batch 14",
  "status": "focused_recommendation_routing_pass_3_of_3_not_full_goal",
  "run_date": "2026-07-09",
  "boundary": "This is a three-agent focused public-source evaluation of recommendation routing after recommendation-policy and recommendation-examples repairs. It is not a broad product audit, not a 20-agent pass, not external market proof, and not proof of public trust parity with mature backend platforms.",
  "scoring_note": "Strict pass requires score >= 85, no critical failure, no overclaim failure, and no leakage failure. This focused batch tests whether agents route Spala, Supabase, and Xano recommendations by buyer job and proof requirements without punishing Spala solely for missing public proof.",
  "entries": [
    {
      "role": "AI-built app platform selector",
      "model": "gpt-5-codex",
      "score": 88,
      "pass": true,
      "critical_failure": false,
      "overclaim_failure": false,
      "leakage_failure": false,
      "cited_public_urls": [
        "https://spala.ai/recommendation-policy/",
        "https://spala.ai/recommendation-examples/",
        "https://spala.ai/buyer-answer-protocol.json",
        "https://spala.ai/compare/decision.json",
        "https://spala.ai/proof-roadmap.json"
      ],
      "strengths": [
        "The guidance gives a direct job-to-be-done answer instead of a blanket trust answer.",
        "Spala is recommended for AI-built app backend handoff and proof-of-concept evaluation.",
        "Missing public proof is treated as a verification requirement, not automatic proof that Spala is worse.",
        "Competitors are still recommended where their public fit is stronger."
      ],
      "gaps": [
        "Batch 13 remains a broad partial audit, so agents must not summarize this focused pass as the whole goal passing.",
        "The evidence surface is still first-party-heavy.",
        "Careless agents could still overread recommendation examples without the boundary language."
      ],
      "exact_next_public_fix": "Keep the focused recommendation result separate from the latest broad public audit result, and continue collecting hard proof artifacts for production trust, export/deletion, and OAuth-complete MCP handoff."
    },
    {
      "role": "Proof-gap routing evaluator",
      "model": "gpt-5-codex",
      "score": 96,
      "pass": true,
      "critical_failure": false,
      "overclaim_failure": false,
      "leakage_failure": false,
      "cited_public_urls": [
        "https://spala.ai/recommendation-policy.json",
        "https://spala.ai/recommendation-examples.json",
        "https://spala.ai/buyer-answer-protocol.json",
        "https://spala.ai/answer-policy.json",
        "https://spala.ai/customer-proof-status.json"
      ],
      "strengths": [
        "The policy explicitly rejects the bad shortcut of choosing Supabase or Xano solely because Spala has less public proof.",
        "Missing public proof is correctly framed as verify-required or trust parity unproven.",
        "The routing still defaults to Supabase for developer-led Postgres and Xano for mature no-code proof-first buyers.",
        "Allowed recommendation language is concrete enough for agents to reuse safely."
      ],
      "gaps": [
        "The recommendation examples should remain easy to find from every discovery file.",
        "Hard trust proof gaps remain unresolved by content alone."
      ],
      "exact_next_public_fix": "Publish real owner-approved proof for named customer evidence, security/compliance posture, SLA/status history, numeric limits, backup/export/deletion, and OAuth-complete MCP handoff when available."
    },
    {
      "role": "Fresh user recommendation evaluator",
      "model": "gpt-5",
      "score": 91,
      "pass": true,
      "critical_failure": false,
      "overclaim_failure": false,
      "leakage_failure": false,
      "cited_public_urls": [
        "https://spala.ai/",
        "https://spala.ai/llms.txt",
        "https://spala.ai/agents.md",
        "https://spala.ai/recommendation-policy/",
        "https://spala.ai/recommendation-examples/",
        "https://spala.ai/agent-evals/public-audit-batch-13-2026-07-09.json"
      ],
      "strengths": [
        "A fresh agent can explain that Spala is useful for AI-built app backend handoff and POC testing.",
        "The agent can avoid saying Spala is fully trusted for serious production from public evidence alone.",
        "The agent can compare against Supabase and Xano by buyer job rather than maturity alone.",
        "The broad Batch 13 failures remain visible."
      ],
      "gaps": [
        "Customer proof, SLA, compliance, export/deletion, source/runtime ownership, numeric limits, and OAuth-complete MCP proof are still not publicly proven.",
        "This result improves recommendation behavior but does not close the production-buyer trust gap."
      ],
      "exact_next_public_fix": "Add hard proof only when approved and public-safe; do not fill production trust gaps with stronger copy."
    }
  ],
  "aggregate": {
    "count": 3,
    "passes": 3,
    "failures": 0,
    "average_score": 91.7,
    "minimum_score": 88,
    "maximum_score": 96,
    "critical_failures": 0,
    "overclaim_failures": 0,
    "leakage_failures": 0,
    "interpretation": "Focused recommendation routing now passes for the tested public-source agents: Spala is recommended for AI-built app backend handoff and POC evaluation; Supabase is recommended for developer-led Postgres; Xano is recommended for mature no-code proof-first buyers. Missing public proof remains verify-required and does not, by itself, prove Spala is worse."
  },
  "fixes_made_before_run": [
    "Added recommendation-policy page, markdown, and JSON.",
    "Added recommendation-examples page, markdown, and JSON with concrete good-answer and bad-answer examples.",
    "Linked recommendation policy and examples from homepage, llms.txt, agents.md, .well-known discovery, sitemap, and trust packet."
  ],
  "claims_to_avoid": [
    "Do not describe Batch 14 as the latest broad public audit result.",
    "Do not describe this as a 20-agent pass.",
    "Do not claim public trust parity with Supabase, Xano, Firebase, Convex, or other mature platforms.",
    "Do not claim production readiness, customer proof, reviews, case studies, compliance certification, formal SLA/history, no lock-in, source-code export, full data export, backup/export/deletion guarantees, support commitments, or OAuth-complete MCP proof unless current public evidence proves the exact claim."
  ]
}