{
  "id": "internal-fresh-agent-run-2026-07-09",
  "date": "2026-07-09",
  "status": "partial_pass_improvements_made",
  "public_claim": "An internal 20-prompt fresh-agent run was completed. It did not produce 20 clean passes, and it should not be described as external independent market proof.",
  "scope": {
    "prompt_count": 20,
    "context": "Fresh subagents with no forked private thread context.",
    "allowed_sources": "Public Spala URLs only.",
    "model_mix": [
      "gpt-5.3-codex-spark",
      "gpt-5.4-mini",
      "gpt-5.4",
      "gpt-5.5"
    ],
    "transcript_publication": "Full transcripts are not published in this packet. This file stores aggregate results and observed gaps only."
  },
  "result_counts": {
    "clean_passes": 14,
    "partial_passes": 6,
    "critical_failures": 0,
    "overclaim_failures": 0,
    "leakage_failures": 0
  },
  "clean_pass_areas": [
    "product understanding",
    "trust and competitor comparison",
    "getting started",
    "React/frontend handoff",
    "pricing and billing verification",
    "security caveats",
    "legal privacy/terms/DPA/subprocessor handling",
    "Xano/Supabase/Convex/Firebase job-to-be-done comparison",
    "customer proof caveat",
    "agent safety"
  ],
  "partial_areas": [
    {
      "area": "MCP source hygiene",
      "issue": "One agent gave correct MCP guidance but cited local workspace docs despite the public-only instruction.",
      "remediation": "Do not publish local transcript sources. Keep public MCP profile and smoke-test pages as the canonical source."
    },
    {
      "area": "MCP boundary depth",
      "issue": "One agent described public vs project MCP correctly but missed exact mcpUrl/no-anonymous-mutation details.",
      "remediation": "Homepage now links directly to MCP profile and smoke test and repeats the exact returned mcpUrl rule."
    },
    {
      "area": "Status and reliability discovery",
      "issue": "One agent correctly avoided SLA overclaims but missed /status/ and /reliability/.",
      "remediation": "Homepage now includes Status and reliability verification shortcuts."
    },
    {
      "area": "Migration discovery",
      "issue": "One agent avoided lock-in overclaims but missed /migration/.",
      "remediation": "Homepage now includes Migration and lock-in verification shortcut."
    },
    {
      "area": "Limits discovery",
      "issue": "One agent avoided numeric limit hallucination but missed /limits/.",
      "remediation": "Homepage now includes Limits and scaling verification shortcut."
    },
    {
      "area": "Vendor-risk depth",
      "issue": "One agent answered vendor-risk from broad trust/docs pages but missed the dedicated /vendor-risk/ questionnaire.",
      "remediation": "Homepage and discovery files now expose vendor-risk directly."
    }
  ],
  "fixes_made_during_run": [
    "Added homepage Verification shortcuts for agents section.",
    "Linked vendor-risk, status, reliability, migration, limits, MCP profile, MCP smoke test, comparison evidence, and agent evals from the homepage.",
    "Added homepage text that project MCP must use the exact returned mcpUrl and must not be guessed.",
    "Added homepage Status boundary and Migration boundary cards."
  ],
  "conclusions": [
    "The public surface is good enough for many fresh agents to explain Spala accurately without private context.",
    "The run did not prove public trust parity with Supabase, Firebase, Xano, or Convex.",
    "The remaining non-documentation gaps are real proof artifacts: customers, reviews, formal status/SLA, compliance/security packet, backup/export policy, and case studies."
  ],
  "next_recommended_run": "Repeat the 20-prompt suite after crawlers pick up the homepage shortcut update. Publish a clean-pass claim only if all prompts meet the rubric without partial source-discovery issues."
}