K-Dense-AI/scientific-agent-skills / skills/hypothesis-generation
hypothesis-generation
VerifiedFormulates evidence-bounded scientific questions, candidate hypotheses, rival explanations, causal or associational claims, discriminating predictions, measurements, and preregistration-ready analysis plans. Used when turning observations or preliminary findings into transparent, testable research plans without treating hypotheses as facts.
assets/falsification_controls_template.json
View on GitHub{
"schema_version": "2.0",
"checklist_id": "FC-SYN-001",
"record_id": "HG-SYN-001",
"hypotheses": [
{
"hypothesis_id": "H1",
"candidate_status": "candidate",
"assumptions": [
"Synthetic assignment, manipulation, and measurements operate as declared."
],
"boundary_conditions": [
"Synthetic eligible units over one prespecified interval."
],
"falsifier": {
"prediction_id": "P1",
"conditions": "The manipulation and measurement checks pass and uncertainty is small enough to distinguish the patterns.",
"observable": "Intermediate M and outcome Y measured at prespecified times.",
"incompatible_result": "Outcome Y changes without the prespecified change in intermediate M.",
"assumption_failure_checks": [
"Verify assignment, manipulation, calibration, missingness, and temporal ordering before interpreting incompatibility."
]
},
"discriminating_tests": [
{
"test_id": "DT1",
"rival_hypothesis_id": "H2",
"focal_expected": "Intermediate M changes before outcome Y with no negative-control reference shift.",
"rival_expected": "Outcome Y and the negative-control reference shift together without the prespecified intermediate change.",
"indeterminate_result": "Both intermediate M and the negative-control reference change, so mixed mechanisms or bias remain possible."
}
],
"nulls": [
{
"null_id": "N1",
"statement": "The population mean of outcome Y is equal under X and comparator at the prespecified horizon.",
"analysis_id": "A1",
"interpretation_limit": "Failure to reject is not equivalence, absence of effect, or proof of the rival."
}
],
"controls": [
{
"control_id": "NC1",
"control_type": "negative_outcome",
"rationale": "The synthetic reference cannot be changed by process M but shares the acquisition workflow.",
"expected_result": "No assignment-linked reference difference after the prespecified acquisition adjustment.",
"failure_implication": "Investigate acquisition, selection, or analysis bias before a causal interpretation."
},
{
"control_id": "PC1",
"control_type": "positive_control",
"rationale": "A known synthetic calibration input checks whether the outcome procedure detects a declared change.",
"expected_result": "The calibration input is recovered within the locked tolerance.",
"failure_implication": "The target test is not interpretable until measurement performance is restored."
}
],
"outcome_interpretation": {
"consistent_with_candidate": "The focal pattern is observed, the rival pattern is not, controls pass, and uncertainty is compatible with the prespecified target.",
"challenges_candidate": "The incompatible result is observed while manipulation, measurements, and declared assumptions remain credible.",
"supports_neither_or_mixed": "The result matches neither pattern, both patterns occur, or an assumption/control fails."
},
"human_review_status": "pending"
},
{
"hypothesis_id": "H2",
"candidate_status": "candidate",
"assumptions": [
"Any synthetic acquisition drift affects the outcome and reference through a shared pathway."
],
"boundary_conditions": [
"Synthetic acquisition periods covered by the declared reference process."
],
"falsifier": {
"prediction_id": "P2",
"conditions": "Independent calibration periods overlap both assigned conditions and reference measurement is valid.",
"observable": "Synthetic negative-control reference and outcome Y by acquisition period.",
"incompatible_result": "The reference remains stable while outcome Y changes reproducibly by assignment.",
"assumption_failure_checks": [
"Verify reference sensitivity, batch overlap, calibration, and missing acquisition records."
]
},
"discriminating_tests": [
{
"test_id": "DT2",
"rival_hypothesis_id": "H1",
"focal_expected": "Reference and outcome shift together by acquisition period.",
"rival_expected": "Outcome changes by assignment without a reference shift and after intermediate M changes.",
"indeterminate_result": "Acquisition period is confounded with assignment, so the rivals cannot be separated."
}
],
"nulls": [
{
"null_id": "N2",
"statement": "The negative-control reference mean is equal across acquisition periods.",
"analysis_id": "A2",
"interpretation_limit": "Failure to reject does not prove absence of drift when precision or reference sensitivity is insufficient."
}
],
"controls": [
{
"control_id": "NC1",
"control_type": "negative_outcome",
"rationale": "The reference is outside process M while sharing the acquisition pathway.",
"expected_result": "A drift explanation predicts a shared acquisition-linked shift.",
"failure_implication": "A stable, sensitive reference challenges the drift candidate."
}
],
"outcome_interpretation": {
"consistent_with_candidate": "Outcome and reference shifts align by acquisition period under valid measurement.",
"challenges_candidate": "The stable reference and independent calibration are incompatible with the declared shared-drift process.",
"supports_neither_or_mixed": "Assignment and acquisition cannot be separated or both mechanism and drift patterns occur."
},
"human_review_status": "pending"
}
]
}
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |