{
  "acceptance_version": "0.1",
  "object_path": "examples/target-product-profiles-for-malaria-diagnostics",
  "manifest": "agent/tools.json",
  "published_results_path": "results.json",
  "notes": "Only one tool here takes a parameter, so only one off-reference call exists. That is a property of the analysis -- two exact tests over a fixed table, each already reporting its own sensitivity variant -- not a gap in the surface.",
  "calls": [
    {
      "call_id": "source-table-published",
      "tool": "source_table",
      "args": {},
      "select": "/n_substantive_criteria",
      "expect": {"kind": "published_pointer", "result_path": "/meta/n_substantive_criteria"},
      "rel_tolerance": 1e-6,
      "abs_tolerance": 1e-9,
      "enabled": true
    },
    {
      "call_id": "t1-tier-composition-published",
      "tool": "tier_composition_test",
      "args": {},
      "select": "/exact_p",
      "expect": {"kind": "published_pointer", "result_path": "/test1_tier_composition/exact_p"},
      "rel_tolerance": 1e-6,
      "abs_tolerance": 1e-9,
      "enabled": true
    },
    {
      "call_id": "t1-presence-absence-sensitivity-published",
      "tool": "tier_composition_test",
      "args": {},
      "select": "/sensitivity_presence_absence",
      "expect": {"kind": "published_pointer", "result_path": "/test1_tier_composition/sensitivity_presence_absence"},
      "rel_tolerance": 1e-6,
      "abs_tolerance": 1e-9,
      "enabled": true,
      "notes": "Collapsing within-criterion repeats removes the pseudoreplication and changes the exact p by 0.012, from 0.592 to 0.580. The null does not depend on the counting convention."
    },
    {
      "call_id": "t2-concordance-published",
      "tool": "concordance_test",
      "args": {},
      "select": "/exact_p",
      "expect": {"kind": "published_pointer", "result_path": "/test2_concordance/exact_p"},
      "rel_tolerance": 1e-6,
      "abs_tolerance": 1e-9,
      "enabled": true
    },
    {
      "call_id": "t2-normalised-sensitivity-published",
      "tool": "concordance_test",
      "args": {},
      "select": "/sensitivity_normalized",
      "expect": {"kind": "published_pointer", "result_path": "/test2_concordance/sensitivity_normalized"},
      "rel_tolerance": 1e-6,
      "abs_tolerance": 1e-9,
      "enabled": true,
      "notes": "Normalising the text changes nothing at all: 14 differing of 23 either way. The differences between the two use-cases are substantive, not whitespace or casing."
    },
    {
      "call_id": "requirement-map-published",
      "tool": "requirement_map",
      "args": {},
      "expect": {"kind": "published_pointer", "result_path": "/descriptive_requirement_map"},
      "rel_tolerance": 1e-6,
      "abs_tolerance": 1e-9,
      "enabled": true
    },
    {
      "call_id": "holm-published",
      "tool": "holm",
      "args": {},
      "select": "/0",
      "expect": {"kind": "published_pointer", "result_path": "/test1_tier_composition/holm_corrected_p"},
      "rel_tolerance": 1e-6,
      "abs_tolerance": 1e-9,
      "enabled": true
    },
    {
      "call_id": "holm-ranked-family-off-reference",
      "tool": "holm",
      "args": {"p_values": [0.01, 0.04, 0.2]},
      "expect": {"kind": "literal", "value": [0.03, 0.08, 0.2]},
      "rel_tolerance": 1e-6,
      "abs_tolerance": 1e-9,
      "enabled": true,
      "notes": "The declared family is two p-values of 0.592 and 0.405, which Holm maps to the same 0.810, so the published call never shows the step-down working. A three-member family with separated p-values does."
    }
  ]
}