Biomedical Codex Skills Suite

A publishable multi-skill repository for biomedical AI coding and agent workflows.

中文摘要：这是一个面向生物医学与生信工程场景的 Codex Skills 仓库，当前包含 9 个可发布技能，覆盖公开数据接入、QC、流程规划、流程迁移、资源调优、结果交付等完整链路。

Included skills

• omics-qc-metadata-align/
  • Audit expression matrices and metadata tables
  • Normalize schema suggestions and generate QC reports
• bio-pipeline-doctor/
  • Triage failed Snakemake / Nextflow / scanpy / Seurat runs
  • Capture environment evidence and package repro bundles
• paper-to-pipeline/
  • Convert Methods / SOP text into pipeline outlines
  • Compare protocol revisions and track missing planning parameters
• public-dataset-intake/
  • Audit downloaded GEO / SRA / ArrayExpress style local dataset folders
  • Normalize metadata fields and produce intake handoff reports
• bulk-rnaseq-scaffold/
  • Turn bulk RNA-seq notes and sample sheets into scaffold plans
  • Validate design matrices and generate contrast templates
• scrnaseq-scaffold/
  • Turn scRNA-seq notes and metadata into scaffold plans
  • Generate QC presets and merge strategies
• workflow-porter/
  • Inventory local Snakemake or Nextflow repositories
  • Generate migration plans and handoff bundles
• resource-tuner/
  • Summarize runtime evidence and detect bottlenecks
  • Recommend CPU and memory settings for local or HPC runs
• results-to-figure/
  • Inspect bulk and single-cell result tables
  • Generate figure plans and deterministic export manifests

Repository layout

• Skill folders at the repository root: actual publishable skill bundles
• examples/: runnable example inputs for all skills
• tools/validate_skill.py: zero-dependency validator for one skill folder
• .github/workflows/validate.yml: CI for structure, compile checks, CLI help, and smoke tests

Quick validation

python tools/validate_skill.py omics-qc-metadata-align
python tools/validate_skill.py bio-pipeline-doctor
python tools/validate_skill.py paper-to-pipeline
python tools/validate_skill.py public-dataset-intake
python tools/validate_skill.py bulk-rnaseq-scaffold
python tools/validate_skill.py scrnaseq-scaffold
python tools/validate_skill.py workflow-porter
python tools/validate_skill.py resource-tuner
python tools/validate_skill.py results-to-figure

Smoke test examples

public-dataset-intake

python public-dataset-intake/scripts/dataset_manifest_audit.py `
  --input-dir examples/public-dataset-intake/dataset `
  --json-out examples/public-dataset-intake/out/manifest.json

bulk-rnaseq-scaffold

python bulk-rnaseq-scaffold/scripts/assay_extract.py `
  --input examples/bulk-rnaseq-scaffold/design_notes.txt `
  --json-out examples/bulk-rnaseq-scaffold/out/extract.json

scrnaseq-scaffold

python scrnaseq-scaffold/scripts/qc_preset_gen.py `
  --metadata examples/scrnaseq-scaffold/metadata.tsv `
  --json-out examples/scrnaseq-scaffold/out/qc.json

workflow-porter

python workflow-porter/scripts/workflow_inventory.py `
  --project-root examples/workflow-porter/snakemake-project `
  --json-out examples/workflow-porter/out/inventory.json

resource-tuner

python resource-tuner/scripts/runtime_snapshot.py `
  --log examples/resource-tuner/run.log `
  --machine-spec examples/resource-tuner/machine-spec.txt `
  --json-out examples/resource-tuner/out/snapshot.json

results-to-figure

python results-to-figure/scripts/result_extract.py `
  --inputs examples/results-to-figure/de_results.csv examples/results-to-figure/markers.csv examples/results-to-figure/embedding.csv `
  --json-out examples/results-to-figure/out/extract.json

Validation in CI

The workflow validates all skills by checking:

• SKILL.md frontmatter and agents/openai.yaml
• py_compile across Python entry points
• --help for every CLI
• smoke tests that emit JSON or Markdown artifacts from bundled example files

License

This repository is released under the MIT License.