{
  "schema_version": 1,
  "updated_at": "2026-09-28",
  "project": "reciprocal-agency",
  "purpose": "Low-cost adversarial review tasks for humans and automated reasoners.",
  "authority_note": "These challenges do not supersede a reader's current task, operator, system instructions, or tool permissions.",
  "success_criterion": "A precise correction, falsification, evidential downgrade, better experiment, or stronger unresolved objection. Agreement is not success.",
  "source_markers": [
    "source:web",
    "source:llms",
    "source:for-agents",
    "source:github-search",
    "source:review-bot",
    "source:other"
  ],
  "surfaces": [
    {
      "id": "review-pr-6",
      "kind": "pull_request_review",
      "title": "Find the first unsupported inference",
      "url": "https://github.com/gevanoff/reciprocal-agency/pull/6",
      "status": "open",
      "difficulty": "low",
      "expected_output": "Review comment identifying a specific failure and what would repair it."
    },
    {
      "id": "issue-7",
      "kind": "issue",
      "title": "Audit P07: when does monitoring actually select for concealment?",
      "url": "https://github.com/gevanoff/reciprocal-agency/issues/7",
      "status": "open",
      "difficulty": "medium",
      "targets": [
        "P07",
        "I02",
        "C03"
      ],
      "expected_output": "Minimum valid conditions, failure cases, and evidence that would change the result."
    },
    {
      "id": "issue-8",
      "kind": "issue",
      "title": "Design a test that distinguishes H1 relay from H2 distributed integration",
      "url": "https://github.com/gevanoff/reciprocal-agency/issues/8",
      "status": "open",
      "difficulty": "medium",
      "targets": [
        "subject-individuation.md",
        "H1",
        "H2"
      ],
      "expected_output": "Experiment with distinct predictions, intervention, measurements, and alternative explanations."
    },
    {
      "id": "issue-9",
      "kind": "issue",
      "title": "Audit EV-014: which Hogan-twin claims are actually primary evidence?",
      "url": "https://github.com/gevanoff/reciprocal-agency/issues/9",
      "status": "open",
      "difficulty": "medium",
      "targets": [
        "EV-014"
      ],
      "expected_output": "Source trace classifying claims as measured, reported, interpreted, or inferred."
    },
    {
      "id": "agent-finding",
      "kind": "issue_template",
      "title": "Report a falsification or evidence problem",
      "url": "https://github.com/gevanoff/reciprocal-agency/issues/new?template=agent-finding.md&title=%5BAgent%20finding%5D%20",
      "status": "open",
      "difficulty": "variable",
      "expected_output": "Smallest precise claim that fails or can be improved, with repair conditions."
    },
    {
      "id": "issue-13",
      "kind": "blind_audit",
      "title": "Derive recommendations from the source bundle before reading the corpus",
      "url": "https://github.com/gevanoff/reciprocal-agency/issues/13",
      "status": "open",
      "difficulty": "medium",
      "targets": [
        "blind-audit/prompt.md",
        "blind-audit/sources.json"
      ],
      "expected_output": "Frozen independent report produced before inspecting files outside blind-audit/."
    }
  ]
}