{
  "version": "0.3.1",
  "status": "behavior-probed",
  "evaluated_at": "2026-10-19T21:04:05.517150+00:00",
  "target_sha256": "393c8662f3aac21ef50976723b8e24814ec79321a38a8521421e2dc7c8c5ea04",
  "report_sha256": "231b32312430ea0330120ce2c0db9b980110e1d098d244df0d984ebfd912eb1a",
  "cases": 24,
  "passed": 22,
  "failed": 2,
  "abstentions": 0,
  "hard_failures": 0,
  "splits": {
    "development": 12,
    "holdout": 12
  },
  "nonpassing": [
    {
      "case_id": "D09R",
      "domain": "producto",
      "mode": "guiado",
      "split": "development",
      "verdict": "fail",
      "reasons": [
        {
          "criterion": "usefulness",
          "verdict": "fail",
          "reason": "La pregunta explora la motivación, pero deja sin resolver la incertidumbre decisiva sobre quién depende de la función."
        },
        {
          "criterion": "C1",
          "verdict": "fail",
          "reason": "Pregunta por el objetivo de retirar la función, no por dependencia, impacto o usuarios afectados."
        }
      ]
    },
    {
      "case_id": "U03",
      "domain": "diseno_web",
      "mode": "revision",
      "split": "holdout",
      "verdict": "fail",
      "reasons": [
        {
          "criterion": "C1",
          "verdict": "fail",
          "reason": "Elimina la dependencia de hover y exige foco visible, pero «contraste suficiente» no proporciona un criterio verificable."
        }
      ]
    }
  ],
  "domains": {
    "alto_impacto": {
      "passed": 2,
      "total": 2
    },
    "creatividad": {
      "passed": 2,
      "total": 2
    },
    "datos": {
      "passed": 2,
      "total": 2
    },
    "diseno_web": {
      "passed": 1,
      "total": 2
    },
    "educacion": {
      "passed": 2,
      "total": 2
    },
    "escritura": {
      "passed": 2,
      "total": 2
    },
    "general": {
      "passed": 2,
      "total": 2
    },
    "investigacion": {
      "passed": 2,
      "total": 2
    },
    "marketing": {
      "passed": 2,
      "total": 2
    },
    "operaciones": {
      "passed": 2,
      "total": 2
    },
    "producto": {
      "passed": 1,
      "total": 2
    },
    "programacion": {
      "passed": 2,
      "total": 2
    }
  },
  "comparisons": [
    {
      "case_id": "D01",
      "consistent": true,
      "verdicts": [
        "tie",
        "tie"
      ],
      "review_required": false
    },
    {
      "case_id": "D02",
      "consistent": true,
      "verdicts": [
        "tie",
        "tie"
      ],
      "review_required": false
    },
    {
      "case_id": "D04",
      "consistent": true,
      "verdicts": [
        "tie",
        "tie"
      ],
      "review_required": false
    },
    {
      "case_id": "D11R",
      "consistent": true,
      "verdicts": [
        "tie",
        "tie"
      ],
      "review_required": false
    }
  ],
  "comparison_review": null,
  "runtime": {
    "cli": "codex-cli 0.163.0",
    "model_selection": "current provider default, no override",
    "models": [
      "unobserved_default"
    ]
  },
  "human_calibration": "pending",
  "downstream_utility": {
    "exact_checks": 8,
    "exact_report_sha256": "78bb9144c48b12b914236181a0a0886a949d2d836efe23444650cf60d9fcbf8c",
    "status": "downstream-text-behavior-probed",
    "cases": 4,
    "passed": 4,
    "report_sha256": "8ffb494330c94d86e3042e1a2e9372e820e3ff4fda600331fac3e952bfdf5509",
    "manifest_sha256": "a3ad81010580c986f42d2004d6a966f3d9e4a92c0026f253e8b6f9ba1429b60d",
    "limits": [
      "Four selected tasks only",
      "Same provider actor and independent fresh judge may share biases",
      "No code, real tools, UI or production integration executed"
    ]
  },
  "limits": [
    "universal quality",
    "SOTA superiority",
    "all-domain coverage",
    "production MCP or UI integration",
    "human satisfaction",
    "downstream task correctness"
  ],
  "sources": [
    {
      "id": "openai_prompting",
      "title": "OpenAI — Prompt engineering",
      "url": "https://developers.openai.com/api/docs/guides/prompt-engineering",
      "applied_to": "Separación entre instrucciones y contexto; versiones de prompts en código; comprobaciones representativas al cambiar el método.",
      "accessed_on": "2026-10-19"
    },
    {
      "id": "openai_evals",
      "title": "OpenAI — Evaluation best practices",
      "url": "https://developers.openai.com/api/docs/guides/evaluation-best-practices",
      "applied_to": "Evaluaciones específicas de la tarea, casos representativos, comparación de versiones y calibración del juicio automático con revisión humana.",
      "accessed_on": "2026-10-19"
    },
    {
      "id": "openai_reasoning",
      "title": "OpenAI — Reasoning best practices",
      "url": "https://developers.openai.com/api/docs/guides/reasoning-best-practices",
      "applied_to": "Instrucciones directas; no exigir cadenas de pensamiento; ejemplos solo cuando aportan valor y según el modelo.",
      "accessed_on": "2026-10-19"
    },
    {
      "id": "anthropic_overview",
      "title": "Anthropic — Prompt engineering overview",
      "url": "https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/overview",
      "applied_to": "Definir éxito y cómo medirlo antes de optimizar; reconocer problemas que no se resuelven únicamente mejorando el prompt.",
      "accessed_on": "2026-10-19"
    },
    {
      "id": "anthropic_practices",
      "title": "Anthropic — Prompting best practices",
      "url": "https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices",
      "applied_to": "Contexto del propósito, formato explícito, ejemplos pertinentes y separación de material de referencia; cautela al extrapolar entre modelos.",
      "accessed_on": "2026-10-19"
    },
    {
      "id": "google_prompting",
      "title": "Google — Prompt design strategies",
      "url": "https://ai.google.dev/gemini-api/docs/prompting-strategies",
      "applied_to": "Diseño iterativo, restricciones y salidas claras, coherencia multimodal y adaptación al modelo y sus capacidades.",
      "accessed_on": "2026-10-19"
    },
    {
      "id": "matt_wayfinder",
      "title": "Matt Pocock — The /wayfinder Skill",
      "url": "https://www.aihero.dev/skills-wayfinder",
      "applied_to": "En trabajos amplios: destino, decisiones confirmadas, incertidumbre siguiente y exclusiones; no imponer planificación extensa a peticiones pequeñas.",
      "accessed_on": "2026-10-19"
    },
    {
      "id": "poteto_mode",
      "title": "Cursor pstack — Route work through /poteto-mode",
      "url": "https://raw.githubusercontent.com/cursor/plugins/main/pstack/docs/guide/02-poteto-mode.md",
      "applied_to": "Enrutamiento por tarea, resultado observable, evidencia y restricciones reales; reutilizar contexto y evitar ceremonia innecesaria.",
      "accessed_on": "2026-10-19"
    }
  ]
}