[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-83052":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":15,"stars7d":15,"stars30d":15,"stars90d":15,"forks30d":15,"starsTrendScore":15,"compositeScore":16,"rankGlobal":9,"rankLanguage":9,"license":17,"archived":18,"fork":18,"defaultBranch":19,"hasWiki":20,"hasPages":18,"topics":21,"createdAt":9,"pushedAt":9,"updatedAt":22,"readmeContent":23,"aiSummary":24,"trendingCount":15,"starSnapshotCount":15,"syncStatus":13,"lastSyncTime":25,"discoverSource":26},83052,"MedSkillOS","albertcheng19\u002FMedSkillOS","albertcheng19","Medical-grade agent skills for clinical and biomedical workflows",null,"Python",77,3,2,1,0,41.81,"MIT License",false,"main",true,[],"2026-06-12 04:01:40","\u003Cdiv align=\"center\">\n\n# 🩺 MedSkillOS\n\n### The open medical skill operating system for AI agents\n\n**Medical-grade agent skills for clinical, biomedical, and scientific workflows.**\n\n\u003Cbr>\n\n[![License: MIT](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-2f80ed?style=for-the-badge)](.\u002FLICENSE)\n![Stage](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FStage-Early%20Development-7c3aed?style=for-the-badge)\n![Agent Native](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FAgent--Native-Skills-10b981?style=for-the-badge)\n![Medical AI](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMedical%20AI-Research%20%26%20Workflow-ef4444?style=for-the-badge)\n![Contributions](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FContributions-Welcome-f59e0b?style=for-the-badge)\n\n\u003Cbr>\n\n**MedSkillOS helps AI agents use medical knowledge, tools, datasets, and research workflows with structure, safety, provenance, and human review.**\n\n\u003Cbr>\n\n[Overview](#-what-is-medskillos) ·\n[Why it matters](#-why-medskillos) ·\n[Architecture](#-architecture) ·\n[Skill Directory](#-skill-directory) ·\n[Install](#-quick-start) ·\n[Contribute](#-contributing) ·\n[Credits](#-acknowledgements--upstream-credit)\n\n\u003C\u002Fdiv>\n\n---\n\n## ✨ What is MedSkillOS?\n\n**MedSkillOS** is an open, agent-native framework for building and running **medical research skills**.\n\nIt is designed for AI agents that need to work across clinical reasoning, biomedical research, evidence synthesis, neurodata processing, bioinformatics, scientific writing, and medical workflow support.\n\nMedSkillOS is not just a folder of prompts.\n\nIt is a standards layer for medical agents:\n\n- reusable **domain skill packs**\n- structured **input\u002Foutput schemas**\n- medical **safety and scope gates**\n- evidence-aware **reasoning artifacts**\n- reproducible **execution traces**\n- quality-control **checklists and evaluations**\n- expert-reviewed **improvement loops**\n\n> **Goal:** turn general AI agents into safer, more useful, more reviewable medical and biomedical research collaborators.\n\n---\n\n## 🧭 Why MedSkillOS?\n\nMedical AI is moving fast. We already have biomedical databases, clinical RAG systems, medical image tools, MCP servers, research assistants, and large collections of AI skills.\n\nBut most systems still optimize for **access**:\n\n| Most tools focus on access to... | MedSkillOS focuses on assurance that... |\n| --- | --- |\n| PubMed, NCBI, FHIR, OMOP, DICOM, ClinicalTrials.gov | the right workflow was selected |\n| guideline documents and biomedical databases | the input was valid and in scope |\n| single-task scripts and utilities | safety gates were applied |\n| generic medical explanations | uncertainty and limitations were stated |\n| one-off prompt templates | provenance and evidence were recorded |\n| large skill catalogs | outputs are reviewable and improvable |\n\nMedSkillOS is built around a simple idea:\n\n> **Medical agents should not only answer. They should show how they worked, what evidence they used, what they are uncertain about, and where human review is required.**\n\n---\n\n## 🧱 Architecture\n\n```mermaid\nflowchart LR\n    U[Clinician \u002F Researcher \u002F Student \u002F Developer] --> A[AI Agent]\n    A --> M[MedSkillOS]\n\n    M --> S[Domain Skill Packs]\n    M --> H[Medical Skill Harness]\n    M --> E[Evidence & Provenance Objects]\n    M --> Q[Quality Gates & Evaluations]\n    M --> R[Expert-Reviewed Refinement]\n\n    S --> D[Diagnostics]\n    S --> N[Clinical Neuroscience]\n    S --> L[Literature & Evidence]\n    S --> B[Bioinformatics & Omics]\n    S --> W[Scientific Writing]\n```\n\nMedSkillOS has four core layers.\n\n### 1. Domain Skill Packs\n\nEach medical domain is organized as a skill pack. A pack may contain agent-readable instructions, schemas, examples, risk boundaries, tests, and reviewer guidance.\n\nInitial focus:\n\n- `diagnostics` — structured clinical reasoning, differential diagnosis, red-flag detection, evidence mapping, and role-specific communication\n- `clinical-neuroscience` — EEG, MEG, fMRI, source localization, spectral analysis, connectivity analysis, and neurophysiology reporting\n\nPlanned and expandable areas:\n\n- literature review and evidence synthesis\n- protocol and study design\n- bioinformatics and omics analysis\n- scientific writing and publication support\n- pharmacology and drug safety\n- radiology and pathology workflows\n- public health and epidemiology\n- medical education\n- medical device and regulatory workflows\n\n### 2. Medical Skill Harness\n\nThe harness runs, validates, audits, and evaluates skills.\n\nIt checks:\n\n- input and output schemas\n- parameter sanity\n- safety boundaries\n- evidence requirements\n- quality-control artifacts\n- provenance metadata\n- deterministic tests\n- regression evaluations\n- human-review requirements\n\nThe goal is not only to run skills, but to make medical-agent workflows **reproducible, inspectable, and improvable**.\n\n### 3. Evidence and Provenance Objects\n\nMedSkillOS standardizes how agents represent evidence, reasoning, data-processing outputs, and review decisions.\n\nCore objects may include:\n\n- `ClinicalQuestion`\n- `EvidenceObject`\n- `SkillRunTrace`\n- `ClinicalExperienceRecord`\n- `ReviewDecision`\n\nThese objects help different skill packs communicate without collapsing everything into unstructured text.\n\n### 4. Expert-Reviewed Self-Refinement\n\nMedSkillOS supports learning from failures, reviewer comments, and user feedback — but not through uncontrolled automatic changes to medical logic.\n\nProposed improvements should pass:\n\n- scope checks\n- safety checks\n- schema validation\n- regression tests\n- expert review\n- versioned promotion\n\nSkill maturity states:\n\n```text\ndraft → candidate → experimental → reviewed → stable\n                                      ↘ deprecated\n```\n\n---\n\n## 📚 Skill Directory\n\nMedSkillOS is designed as a growing medical skill registry. The directory should stay readable: the README highlights major domains, while the full catalog can live in `\u002Fdocs`, `\u002Fskills`, or generated index files.\n\n### 🩺 Diagnostics\n\nStructured clinical reasoning skills that help agents reason clearly, surface missing information, and communicate safely.\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Example skills\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n- `problem-representation`\n- `red-flag-detection`\n- `differential-diagnosis-builder`\n- `evidence-for-against-mapper`\n- `missing-information-identifier`\n- `source-router`\n- `evidence-grader`\n- `doctor-summary`\n- `nurse-handoff`\n- `patient-explanation`\n- `feedback-classifier`\n- `reviewer-gate`\n\n\u003C\u002Fdetails>\n\n### 🧠 Clinical Neuroscience\n\nReproducible workflows for neurodata processing and reporting.\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Example skills\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n- `validate-bids-dataset`\n- `load-eeg-meg-raw`\n- `inspect-raw-quality`\n- `apply-notch-filter`\n- `apply-bandpass-filter`\n- `detect-bad-channels`\n- `fit-ica-or-ssp`\n- `generate-eeg-meg-qc-report`\n- `run-fmriprep-wrapper`\n- `inspect-fmriprep-outputs`\n- `compute-psd`\n- `compute-time-frequency`\n- `run-source-localization`\n- `compute-connectivity`\n- `generate-neuro-report`\n\n\u003C\u002Fdetails>\n\n### 🔍 Literature & Evidence\n\nSkills for finding, screening, appraising, and synthesizing biomedical literature.\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Example skills to add\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n- biomedical search strategy builder\n- PubMed query optimizer\n- high-value paper screener\n- evidence map generator\n- contradiction resolver\n- claim-to-paper verifier\n- research gap finder\n- reporting guideline matcher\n\n\u003C\u002Fdetails>\n\n### 🧪 Protocol & Study Design\n\nSkills for turning research questions into executable medical research plans.\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Example skills to add\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n- aim and hypothesis designer\n- cohort protocol planner\n- inclusion\u002Fexclusion criteria builder\n- endpoint definition assistant\n- sample size and power planner\n- real-world evidence study designer\n- biomarker validation strategy designer\n- feasibility-aware study planner\n\n\u003C\u002Fdetails>\n\n### 🧬 Bioinformatics & Omics\n\nSkills for reproducible analysis of molecular and biomedical datasets.\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Example skills to add\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n- differential expression analysis\n- batch effect correction\n- GO\u002FKEGG enrichment\n- GSEA and GSVA\n- WGCNA\n- immune infiltration analysis\n- survival modeling\n- ROC and diagnostic performance\n- single-cell analysis planning\n- multi-omics integration\n\n\u003C\u002Fdetails>\n\n### ✍️ Scientific Writing\n\nSkills for transforming research work into clearer scientific communication.\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Example skills to add\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n- abstract builder\n- method section writer\n- results narrative builder\n- discussion architect\n- medical English precision editor\n- journal matcher\n- cover letter drafter\n- reviewer response planner\n- reporting guideline compliance checker\n\n\u003C\u002Fdetails>\n\n---\n\n## 🗂️ Recommended Repository Layout\n\n```text\nMedSkillOS\u002F\n  README.md\n  LICENSE\n  NOTICE.md\n  CONTRIBUTING.md\n\n  skills\u002F\n    diagnostics\u002F\n      problem-representation\u002F\n        SKILL.md\n        skill.yaml\n        schemas\u002F\n        examples\u002F\n        tests\u002F\n        risk.md\n\n    clinical-neuroscience\u002F\n      validate-bids-dataset\u002F\n        SKILL.md\n        skill.yaml\n        schemas\u002F\n        examples\u002F\n        tests\u002F\n        risk.md\n\n  docs\u002F\n    catalog.md\n    architecture.md\n    safety-model.md\n    contribution-guide.md\n    third-party-notices.md\n\n  evals\u002F\n    cases\u002F\n    rubrics\u002F\n    regression\u002F\n\n  schemas\u002F\n    ClinicalQuestion.schema.json\n    EvidenceObject.schema.json\n    SkillRunTrace.schema.json\n    ReviewDecision.schema.json\n```\n\nA skill should define:\n\n- what it does\n- when to use it\n- when **not** to use it\n- required inputs\n- expected outputs\n- safety boundaries\n- evidence requirements\n- quality-control requirements\n- provenance requirements\n- known failure modes\n\n---\n\n## 🚀 Quick Start\n\n> MedSkillOS is currently in early development. Directory names and install paths may change as the project matures.\n\nClone the repository:\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Falbertcheng19\u002FMedSkillOS.git\ncd MedSkillOS\n```\n\nInstall selected skills into your agent framework:\n\n```bash\n# Example: install all available skills into a local agent skill directory\nmkdir -p ~\u002F.local\u002Fshare\u002Fagent-skills\ncp -r skills\u002F* ~\u002F.local\u002Fshare\u002Fagent-skills\u002F\n```\n\nFor OpenClaw-style skill loading:\n\n```bash\nmkdir -p ~\u002F.openclaw\u002Fskills\ncp -r skills\u002F* ~\u002F.openclaw\u002Fskills\u002F\n```\n\nFor Claude-style local skills:\n\n```bash\nmkdir -p ~\u002F.claude\u002Fskills\ncp -r skills\u002F* ~\u002F.claude\u002Fskills\u002F\n```\n\nThen ask your agent:\n\n```text\nWhat MedSkillOS skills are available, and when should each one be used?\n```\n\n---\n\n## 🧪 Example Usage\n\n```text\nUse the differential-diagnosis-builder skill.\n\nPatient summary:\n- 45-year-old with new headache and transient visual symptoms\n- no fever\n- history of hypertension\n\nTask:\nCreate a structured differential diagnosis, identify red flags,\nlist missing information, and clearly state when urgent clinical review is needed.\n```\n\nExpected MedSkillOS-style output:\n\n```text\n1. Problem representation\n2. Red flags and immediate safety concerns\n3. Differential diagnosis with evidence for\u002Fagainst\n4. Missing information\n5. Suggested source routing\n6. Uncertainty and limitations\n7. Human review requirement\n```\n\n---\n\n## 🛡️ Safety and Scope\n\nMedSkillOS is designed for:\n\n- medical research\n- biomedical data processing\n- clinical workflow support\n- medical education\n- expert-reviewed agent development\n- reproducible scientific workflows\n\nMedSkillOS is **not**:\n\n- a replacement for clinicians\n- a diagnostic authority\n- a treatment recommendation engine\n- a scraped medical textbook\n- a guideline mirror\n- a general medical chatbot\n- a marketplace of unverified tools\n\nMedical outputs generated with MedSkillOS require appropriate human review.\n\n---\n\n## ✅ Quality Model\n\nEvery mature skill should pass two layers of review.\n\n### Skill Quality\n\n- clear trigger conditions\n- explicit non-use cases\n- structured input\u002Foutput contract\n- testable behavior\n- reliable examples\n- safe tool usage\n- reproducible outputs\n\n### Medical Quality\n\n- scope boundaries\n- uncertainty reporting\n- evidence awareness\n- clinical safety warnings\n- source provenance\n- reviewer handoff\n- no unsupported medical authority claims\n\n---\n\n## 🧑‍🔬 Contributing\n\nMedSkillOS welcomes contributions from:\n\n- physicians\n- nurses\n- pharmacists\n- clinical neuroscientists\n- radiologists\n- pathologists\n- genetic counselors\n- biomedical researchers\n- medical students\n- patients and caregivers\n- software engineers\n- evaluation designers\n- safety and governance reviewers\n\nYou do not need to write code to contribute. Valuable contributions include:\n\n- workflow designs\n- skill drafts\n- examples\n- failure cases\n- evaluation rubrics\n- safety boundaries\n- source-routing rules\n- domain review comments\n- documentation improvements\n\nSuggested contribution flow:\n\n```text\n1. Propose a skill or improvement\n2. Define scope and non-scope\n3. Add examples and expected outputs\n4. Add safety and evidence requirements\n5. Add tests or evaluation cases\n6. Request review\n```\n\n---\n\n## 🧩 Acknowledgements & Upstream Credit\n\nMedSkillOS builds on ideas, workflows, and open-source work from the medical AI skills community.\n\nWe gratefully acknowledge:\n\n- [Aperivue \u002F medsci-skills](https:\u002F\u002Fgithub.com\u002FAperivue\u002Fmedsci-skills)\n- [FreedomIntelligence \u002F OpenClaw-Medical-Skills](https:\u002F\u002Fgithub.com\u002FFreedomIntelligence\u002FOpenClaw-Medical-Skills)\n- [AIPOCH \u002F medical-research-skills](https:\u002F\u002Fgithub.com\u002Faipoch\u002Fmedical-research-skills)\n\nSome skills, categories, workflow patterns, or documentation ideas in MedSkillOS may be derived from, adapted from, or inspired by these upstream projects.\n\nWhen upstream content is copied or adapted:\n\n- keep original copyright notices\n- retain original license text where required\n- document the source repository\n- note meaningful modifications\n- do not import files with unclear or incompatible licensing\n- respect third-party content restrictions inside upstream repositories\n\nRecommended notice file:\n\n```text\ndocs\u002Fthird-party-notices.md\n```\n\n---\n\n## 📄 License\n\nMedSkillOS is licensed under the [MIT License](.\u002FLICENSE).\n\nThird-party content, adapted skills, bundled checklists, datasets, scripts, and examples may be subject to their own licenses. Their original licenses and attribution notices must be preserved.\n\n---\n\n## ⚕️ Medical Disclaimer\n\nMedSkillOS is for research, education, workflow support, and expert-reviewed agent development.\n\nIt is not a validated clinical tool and must not be used as a replacement for qualified medical judgment, diagnosis, or treatment.\n\nAlways involve qualified professionals for clinical decisions.\n\n---\n\n\u003Cdiv align=\"center\">\n\n### Build safer medical agents. Share better research workflows. Make every skill reviewable.\n\nIf MedSkillOS helps your work, consider starring the repository ⭐\n\n\u003C\u002Fdiv>\n","MedSkillOS 是一个面向临床和生物医学工作流的医疗级代理技能操作系统。该项目通过Python语言构建，旨在为AI代理提供结构化、安全且可追溯的医疗知识、工具、数据集及研究工作流程支持。其核心功能包括领域技能包复用、输入输出模式标准化、医疗安全与范围控制门、基于证据的推理工件以及可重复执行跟踪等。此外，MedSkillOS还强调了对输出结果的质量控制检查与评估机制，并鼓励专家参与改进循环以持续优化系统性能。该平台特别适用于需要在临床推理、生物医学研究、科学写作等领域内工作的AI助手场景，确保这些智能体能够更安全、更有用地参与到医疗与生物医学研究合作中去。","2026-06-11 04:10:01","CREATED_QUERY"]