{"id":16872,"plugin_id":"plugins_6a6a29948a7c8191ace6787d5ae074bb","kind":"skill","collection_source":null,"comparison_source":null,"observed_at":"2026-09-30T23:13:50.715Z","digest":"850fd9c0afd08e21d25b8dfdec1373e2a0d95fcc820fbc54447326f89403db4d","against":null,"payload":{"name":"jinko-task-trial-data-scoping","description":"Find and shortlist ClinicalTrials.gov registry and posted-results records for biomedical modeling evidence. Use for NCT discovery, status/phase/results screening, endpoint and population inventory, comparator landscapes, and ongoing-trial intelligence. Do not use for PubMed publication discovery, quantitative extraction, protocol authoring, Jinkō trial execution, calibration, model building, or systematic reviews.","included_files":[{"relative_path":"agents/openai.yaml","size_in_bytes":288},{"relative_path":"assets/shortlist-schema.json","size_in_bytes":1911},{"relative_path":"evals/evals.json","size_in_bytes":3527},{"relative_path":"evals/files/trial-scoping-use-cases.md","size_in_bytes":2887},{"relative_path":"scripts/clinical_trials.py","size_in_bytes":9240},{"relative_path":"scripts/common.py","size_in_bytes":658},{"relative_path":"scripts/compile_trials.py","size_in_bytes":3765},{"relative_path":"scripts/validate_shortlist.py","size_in_bytes":2804}],"skill_md_contents":"---\nname: jinko-task-trial-data-scoping\ndescription: >-\n  Find and shortlist ClinicalTrials.gov registry and posted-results records for\n  biomedical modeling evidence. Use for NCT discovery, status/phase/results\n  screening, endpoint and population inventory, comparator landscapes, and\n  ongoing-trial intelligence. Do not use for PubMed publication discovery,\n  quantitative extraction, protocol authoring, Jinkō trial execution,\n  calibration, model building, or systematic reviews.\ncompatibility: Requires network access. ClinicalTrials.gov v2 needs no credentials.\nmetadata:\n  author: Nova In Silico\nlicense: MIT\n---\n\n# Clinical Trial Data Scoping\n\nProduce a registry candidate inventory, not extracted endpoint data. A registered\noutcome is not evidence that numeric results were posted.\n\n## Frame\n\nEstablish the condition, intervention or mechanism class, population, comparator,\nand purpose: endpoint availability, control-arm or natural-history evidence,\ndose/regimen context, safety, or development-landscape intelligence. Confirm any\nrequired status, phase, study type, posted-results requirement, and shortlist\nsize. Do not impose a status filter silently.\n\nReuse the Entity Table from `jinko-task-literature-search` when available;\notherwise build `canonical_name`, `synonyms`, `mesh_term`, `related_entities`,\n`intent_groups` (`Data`), and `exclusions`. Confirm consequential aliases and\nfilters before network calls.\n\n## Search\n\n1. Build distinct ClinicalTrials.gov angles from the relevant facets:\n   intervention aliases, condition aliases, mechanism class, comparator or\n   standard of care, and population/outcome. Prefer precise terms over one broad\n   query.\n2. Run `scripts/clinical_trials.py` once per angle with separate output files.\n   Use `--status`, `--phase`, and `--require-results` only when required by the\n   approved frame. The script owns API filtering, raw-response persistence, and\n   normalized registry fields.\n3. Run `scripts/compile_trials.py` over the angle outputs. It deduplicates by NCT\n   ID, preserves query provenance, and ranks by angle count, posted-results\n   availability, and record completeness.\n\nThis is one search pass. Broader mechanism, sponsor, country, site, or comparator\nqueries are separate user-approved follow-ups.\n\n## Shortlist\n\nInspect each retained record against the stated purpose. Distinguish:\n\n- registry-only design or recruitment metadata;\n- posted ClinicalTrials.gov results;\n- a publication linked to an NCT identifier.\n\nSet `verification_passed` only when the record contains purpose-relevant signals,\nsuch as a matching population/intervention, specified outcome and timeframe,\nenrollment and eligibility, appropriate design, or the required results modules.\nState the observed signals in `verification_note`; do not infer numeric endpoint\navailability from `hasResults` alone.\n\nComplete the fields in `assets/shortlist-schema.json` and run\n`scripts/validate_shortlist.py` before presenting `shortlist.json`. Trial records\nuse `intent_group = Data`, an appropriate registry/results `evidence_type`, and\n`nct_id` as their primary identifier.\n\nWhen associated publications are needed, pass selected NCT IDs to\n`jinko-task-literature-search` as `<NCT_ID>[si]` angles. Use\n`jinko-task-extract-data-table` only after a quantitative source has been\nidentified and inspected.\n\n## Artifacts\n\n- `frame.json`: approved scope, Entity Table, filters, and angle definitions;\n- per-angle normalized JSON, raw API JSON, and table JSON;\n- `merged_trials.json`: deterministic cross-angle candidate pool;\n- `shortlist.json`: schema-valid prioritized candidates.\n\nPresent a concise Markdown view with NCT link, title, phase/status, results\navailability, population, interventions, primary outcomes, and priority rationale.\nClearly separate scoped candidates from analysis-ready data.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}