Update to NGS Analysis Workbench

{
  "name": "design-ngs-analysis",
  "description": "Translate a biological question or scientific outcome into a defensible NGS analysis plan. Use to define experimental units, groups, covariates, contrasts, endpoints, references, validity gates, evidence needs, methods, or supportable claims before choosing an executable workflow.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 299
    }
  ],
  "skill_md_contents": "---\nname: design-ngs-analysis\ndescription: Translate a biological question or scientific outcome into a defensible NGS analysis plan. Use to define experimental units, groups, covariates, contrasts, endpoints, references, validity gates, evidence needs, methods, or supportable claims before choosing an executable workflow.\n---\n\n# Design NGS Analysis\n\nRead [AnalysisContext](../../references/analysis-context.md). Design backward\nfrom the scientific decision, not forward from a pipeline. If material identity\nor relationships are insufficient, follow\n[understand-ngs-data](../understand-ngs-data/SKILL.md) first. This skill is\nread-only and does not register or execute a workflow.\n\n## Resolve the scientific model\n\nEstablish or leave explicitly unknown:\n\n- scientific question, decision, assay, and current data state\n- biological experimental unit versus technical partitions\n- groups, pairing, repeated measures, batches, covariates, and confounding\n- contrasts, baselines, endpoints, and quantification level\n- organism, genome build, reference, annotation, and identifiers\n- evidence required for the claim, failure conditions, and unsupported claims\n\nDo not treat cells, reads, lanes, or capture channels as biological replicates.\nLeave unidentifiable comparisons untested. Keep model-based batch adjustment\ndistinct from visualization-oriented batch removal.\n\nLoad only the applicable scientific reference:\n\n- [Basic FASTQ QC](../../references/fastq-qc.md)\n- [Bulk RNA-seq](../../references/bulk-rnaseq.md)\n- [Single-cell RNA-seq](../../references/single-cell-rnaseq.md)\n- [Detailed single-cell guidance](../../references/single-cell-qc-annotation-umap-heuristics.md), before choosing cell thresholds, correction, annotation, or embedding policy\n- [Demultiplexing](../../references/bcl-demultiplexing.md), for run-folder and sample-assignment endpoints\n- [DNA variants](../../references/dna-variants.md), for germline, somatic, or UMI-panel endpoints\n- [Epigenomics](../../references/epigenomics.md), for accessibility or antibody-targeted endpoints\n- [Microbiome](../../references/microbiome.md), for amplicon or shotgun taxonomy/function endpoints\n\nFor differential expression, apply the bulk or single-cell reference to the\nactual matrix state, donor-level replication, design, and requested contrast.\nDo not restart read processing when suitable counts and metadata already exist.\n\nFor each endpoint define validated inputs, method assumptions, validity gates,\nmeasurements and denominators, comparison structure, required QC/provenance,\nand criteria for supported, unsupported, or inconclusive claims. Catalogs may\nestablish feasibility, but installed software does not choose the scientific\nmethod.\n\n## Output: AnalysisPlan\n\nAdd an `analysis_plan` artifact with:\n\n1. objective and decision\n2. evidence-backed starting point\n3. scientific model and reference assumptions\n4. endpoints, comparisons, methods, and rationale\n5. evidence contract and validity gates\n6. limitations, unknowns, and unresolved choices\n7. supported, unsupported, and conditional claims\n8. implementation requirements without readiness or approval claims\n\nIf results change the question, supersede the prior plan and preserve why.\n"
}