informed-patient is a Claude Code plugin with 1 hand-picked skill, indexed on Flowy in the Research category.
> /plugin marketplace add drcathicks/informed-patient> /plugin install informed-patient@informed-patient
This Claude Skill provides a framework to help you use Claude to research medical evidence to inform specific health questions, primarily designed around use cases like trying to understand and organize questions about symptoms, preparing for an upcoming medical appointment, and gathering evidence for either differential diagnosis questions or condition progression. Users have also been successful at using these instructions as structured prompts or workflows with any AI model, including local models.
When you ask a medical question, Claude will walk you through a semi-structured cognitive interview to elicit context for its search, then conduct a search following a set of guidelines for evidence robustness. The skill provides two potential search "modes": by default it uses a more structured search across a defined evidence source hierarchy, but can use an open search mode for scenarios where you raise a more complex questions that falls between conditions, when you want a more exploratory mode, or when the opinionated search results are thin.
Alongside searching and synthesizing sources, Claude will weigh the strength of evidence or relevant concerns according to a set of evidence "red flags." The results will be shared in a "health evidence review," including sources, search terms, and evidence red flags to be aware of, and questions generated from the exercise to bring to your healthcare team.
The aim of this skill is not to replace your healthcare team, but to guide your use of AI as a structured search tool around health information, help you advocate for yourself and your needs, and help you become an "informed patient."
Nothing here is intended as medical advice. This skill and all of the text in this repo is provided in order to give you an example to work from, and in order to provide users with suggestions for technical help in navigating how language models interact with the scientific literature and web information about general health topics. Do not make health decisions based on this data or any output from this skill without consulting a qualified clinician.
This repository provides two installation methods depending on your Claude client:
This repository is structured as a Claude Code plugin. To install, run these commands in Claude Code:
/plugin marketplace add https://github.com/DrCatHicks/informed-patient.git
/plugin install informed-patient@informed-patient
For more on Claude Code plugins, see the plugin documentation.
For Claude Desktop or Claude Web, install via the Settings UI:
informed-patient-skill.zip from the latest releaseYou may want to download your own health records, which can significantly improve the utility of this skill. Nothing in this skill requires a health record, but if you wish to do this, JT Perry has created a tool to make it easier for individuals to download their health records (which in the US, you have a right to access because of the 21st Century Cures Act (Pub. L. No. 114-255). You can see his tool here, or see the repo. Please review the warnings and think carefully about how you feel comfortable accessing and using this sensitive information, as you will be solely responsible for the risks.
You may also wish to download your health record in order to analyze it yourself, entirely separately from AI, and filter details in it for your use of this skill.
This skill is opt-in — it shouldn't activate automatically when you ask health questions. To use it, explicitly tell Claude you'd like to use the informed-patient skill, then describe your situation. Modify trigger instructions to change this.
Claude will ask questions before searching. This interview is the foundation for the evidence review. The more specific and accurate your inputs, the more useful the search will be.
This is a thinking tool, not a diagnostic tool.
Using AI for health is consistently one of the top use cases documented for emerging AI tools. Many people turn to AI for help when trying to make sense of a diagnosis, prepare for a specialist appointment, or understand what the research actually says about a condition. People also struggle with recognized risks of using AI, such as inaccurate results, encouragement of overly confirmatory thinking, and lack of source clarity.
The Informed-Patient Skill runs a structured process to help you use Claude with evidence science-informed instructions to create an "evidence review" around a specific health question. Using this Skill, Claude will dialogue with you to create a guided symptom/or medical scenario overview, will design a literature search according to a structured evidence hierarchy, and will create a structured document you can bring to your next medical appointment or use to document your questions.
At the end of the exercise, output is saved in a Health Evidence Review that includes:
It is impossible to anticipate every possible medical scenario. This Skill has complex instructions, and is meant to provide useful instructions to Claude, but the quality and depth of the information you give it will change how specific and helpful its search will be.
You can experiment with adapting this skill to your needs. You may want to consider:
This skill is opinionatedly designed to use evidence guardrails in the background, as well as take advantage of Claude's ability to help you create your own personalized patient inventory. It is also structured to push Claude against speculating or "fill in the answer" defaults, to be honest about what it found, what it couldn't find, to provide evidence for how much confidence you should have in competing hypotheses, as well as suggest directions that would confirm or deny working hypotheses & scenarios.
My aim was to provide a practical thinking tool that strikes a balance between giving you assistance that lightens the load in searching complex literature, but also gives clarity and directions for investigation. You are centered as the decision-maker: you control the symptom description, rank questions, and have access to records for further searching or your own cross-verification. I wanted to create a tool that would help people who aren't researchers, scientists or doctors, and who are dealing with something hard, and who may have limited time and energy to wade through complex medical details. Therefore, you can expand or contract your use of the Skill; a quick session of 10 minutes is enough for a review, but you can also run multiple reviews in a row to continue investigating.
The references give Claude a structured set of instructions for assessing and flagging evidence red flags, such as missing evidence, understudied conditions, or barriers in care. These are currently set as background references, but you may wish to read and modify them yourself.
Thinking about your health is high stakes, and we are often cognitively biased toward wanting to find a single answer. To help you get better answers, this Skill is opinionated about holding a high standard of evidence and instructs Claude to try to counter some of the aggressively confirmatory patterns in AI chatbots used for health. In the case of trying to explain symptoms, Claude should explicitly push you to consider multiple hypotheses and evaluate the strength of evidence, as well as generate suggestions of evidence that would strengthen or weaken confidence in the hypothesis.
This Skill is meant to help you organize your experience and evaluate evidence. It should not be used to tell you what you have or what to do about it.
While I have continually tested the sources this Skill produces and refined the Skill across many disparate searches, it is always possible for AI to be inaccurate.
You can use the EVALUATION.md file as a starting place to check that this Skill is performing as expected across different scenarios.
The aim of this skill is not to provide "the answer" or end of a question, but to sharpen your investigatory skills and give you support to advocate for yourself as a patient or on behalf of a patient.
In keeping with that vision, in this Skill, I have used opinionated evidence-based principles to design for eliciting diverse and functionally-relevant descriptions of symptomology, for assessing the clinical literature, and for keeping in mind the complex causality of healthcare decision-making. Going through an iterative process (intake, search for sources, source summaries, evidence red flag evaluation, synthesis) introduces step-by-step thinking and more reflection, while still letting you take advantage of AI assistance. Running multiple reviews will strengthen this process further.
View @drcathicks’s profileThis skill guides you through semi-structured, interactive goal-setting using the technique of Mental Contrasting with Implementation Intentions (MCII), an evidence-based psychological exercise that draws on a self-regulation strategy to improve learning
@drcathicksBuild your expertise, not just your projects. This skill uses an adaptive "dynamic textbook" approach to help you integrate science-based expertise building exercises while doing agentic coding.
@drcathicksinformed-patient is a Claude Code plugin with 1 hand-picked skill, indexed on Flowy in the Research category. Install it with the command on its page. It includes informed-patient. Its skills do not fire on their own yet. Request auto-invocation to have Flowy route them as you prompt. Free and open source.
Is this plugin yours?
Claim it with GitHubSubmit a pluginPromote it