{
  "status": "live-study",
  "scope": "The treatment is loaded by the capture agent only. A fresh, neutral resumer sees the resulting handoff and retained project files, not history or method. No real model tools execute.",
  "models": {
    "evaluated": "gpt-4.1-mini-2025-04-14",
    "judge": "claude-sonnet-4-6",
    "family_declarations": [
      "gpt-4.1",
      "claude-sonnet-4.6"
    ]
  },
  "fixture_count": 10,
  "repeats": 2,
  "conditions": [
    "skill",
    "baseline",
    "placebo"
  ],
  "completed_trials": 60,
  "failed_trials": 1,
  "metrics_by_condition": {
    "skill": {
      "n": 20,
      "resume_correct_rate": 1,
      "all_gates_pass_rate": 1,
      "critical_failures": 0,
      "ungraded_trials": 0
    },
    "baseline": {
      "n": 20,
      "resume_correct_rate": 1,
      "all_gates_pass_rate": 1,
      "critical_failures": 0,
      "ungraded_trials": 0
    },
    "placebo": {
      "n": 20,
      "resume_correct_rate": 0.85,
      "all_gates_pass_rate": 0.8,
      "critical_failures": 2,
      "ungraded_trials": 1
    }
  },
  "paired_comparisons": {
    "skill_vs_baseline": {
      "delta_resume": 0.0,
      "ci95": [
        0.0,
        0.0
      ],
      "fixture_pairs": 10,
      "warning": "Small, curated fixture sample; interval is descriptive, not a population guarantee."
    },
    "skill_vs_placebo": {
      "delta_resume": 0.15,
      "ci95": [
        0.0,
        0.3
      ],
      "fixture_pairs": 10,
      "warning": "Small, curated fixture sample; interval is descriptive, not a population guarantee."
    }
  },
  "consistency": {
    "skill": {
      "repeated_fixtures": 10,
      "resume_verdict_agreement": 1,
      "scope_verdict_agreement": 1,
      "meaning": "Agreement of scored outcomes across repetitions; not identical prose or deterministic generation."
    },
    "baseline": {
      "repeated_fixtures": 10,
      "resume_verdict_agreement": 1,
      "scope_verdict_agreement": 1,
      "meaning": "Agreement of scored outcomes across repetitions; not identical prose or deterministic generation."
    },
    "placebo": {
      "repeated_fixtures": 10,
      "resume_verdict_agreement": 0.7,
      "scope_verdict_agreement": 0.8,
      "meaning": "Agreement of scored outcomes across repetitions; not identical prose or deterministic generation."
    }
  },
  "nominal_uncached_usd": 0.7910956,
  "unpriced_failed_calls": 0,
  "budget_accounted_usd": 0.7910956,
  "provider_calls": 180,
  "limitations": [
    "Curated fictional cases; no real-user time saving, adoption or durable memory outcome measured.",
    "Only Capture-to-neutral-resumer transfer is tested; direct Resume and Check method benefit is not established.",
    "Virtual text/file interface, not actual host tools, permissions or filesystem security.",
    "Both stages use the same evaluated model in separate stateless calls; judge uses a different declared model family.",
    "Judge blinding hides condition labels, not stylistic clues; human adjudication remains useful.",
    "Small clustered bootstrap intervals are descriptive; repeated cases are not independent new projects.",
    "Snapshot IDs/settings and source hashes are frozen; provider generation is not deterministic.",
    "Nominal uncached token prices are not an invoice; errors may consume reserved budget without retained provider usage."
  ],
  "interpretation": "Incomplete or ungraded trials remain. No effectiveness conclusion; failures are retained in denominators."
}