What does this skill do, and when should you use it?
This skill uses a staged workflow to develop clinical trial protocols for drugs or medical devices. It researches similar trials, FDA databases, regulatory pathways, and published protocol guidance before producing a research summary or a complete protocol draft. It supports Research Only and Full Protocol modes and stores progress in waypoint files for resumption. The workflow requires a Clinical Trials MCP server, while later statistical calculations require Python, SciPy, and NumPy.
Collects intervention details and preserves supplied technical or research context; searches similar trials through the Clinical Trials MCP server; accesses FDA 510(k), PMA, De Novo, Drugs@FDA, Orange Book, and Purple Book databases; generates research summaries, protocol sections, sample-size calculations, and Markdown protocol documents; writes metadata, research findings, and section files to the waypoints/ directory and resumes from completed steps.
- A clinical research team evaluating a new medical device and researching comparable trials and FDA pathways.
- A drug-development team preparing preliminary protocol material for an investigational product submission.
- A researcher who wants a focused clinical-trial research summary before committing to full protocol generation.
- A protocol team developing study design, intervention, assessment, statistics, and operations sections in stages.
How do you install this skill?
- This is a high-risk clinical and regulatory drafting aid and should not be used directly as a clinical protocol or regulatory submission.
- Execution depends on the ClinicalTrials.gov MCP, Python dependencies, and scripts/files not fully evidenced here; unavailable services, path mismatches, or missing dependencies may prevent completion.
- Research findings, regulatory references, statistical calculations, participant privacy, and data flows should be independently checked by qualified clinical, statistical, regulatory, IRB, and legal reviewers.
- Chinese-language support and reachability of external services from mainland-China networks are not documented.
- Shell / CLI
- Network access
- Local filesystem
- MCP Server
Pythonscipy >= 1.11.0numpy >= 1.24.0Clinical Trials MCP server
Install the healthcare plugin collection:
/plugin marketplace add anthropics/healthcare
/plugin install healthcare@healthcare
The SKILL.md also requires a Clinical Trials MCP server and describes installing it by dragging an .mcpb file into Claude Desktop or configuring it manually in Claude Desktop settings. The README lists Clinical Trials as a hosted MCP server. The documented Python dependency command is: pip install -r requirements.txt.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/anthropics/healthcare.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/plugins/healthcare/skills/clinical-trial-protocol" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Create a clinical trial protocol,
- Generate protocol for [device/drug],
- Help me design a clinical study,
- Research similar trials for [intervention].
Invoke the skill and select 1 (Research Only), 2 (Full Protocol), or 3 (Exit). Example prompts include “Create a clinical trial protocol,” “Generate protocol for [device/drug],” “Help me design a clinical study,” and “Research similar trials for [intervention].” Research Only runs initialization and research; Full Protocol continues through protocol generation and pauses before final concatenation for review.
What are this skill's strengths and limitations?
- Covers protocol development for both medical devices and drugs.
- Combines similar-trial research with FDA database and regulatory-pathway research.
- Saves intermediate work in waypoint files and supports resuming interrupted workflows.
- Includes a validated sample-size calculation script and Markdown deliverables.
- The Clinical Trials MCP server is required for Step 1, with no fallback if it is unavailable.
- Python, SciPy, and NumPy are required for the statistical calculation stage.
- The output is not FDA or IRB approval and does not replace biostatistical, clinical, regulatory, or legal review.
- The source provides no test suite, detailed platform matrix, or complete example final protocol.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Clinical Trial Protocol Builder this page ✓ Anthropic · Official | 41 · Not recommended | ★ 425 | 1mo ago | — |
| FHIR Clinical Data Connector ✓ Anthropic · Official | 49 · Use with care | ★ 425 | 1mo ago | — |
| VSS Video Analytics Query ✓ NVIDIA · Official | 54 · Use with care | ★ 3.6k | 11d ago | Apache-2.0 |
| BigQuery Asset Impact Analysis ✓ Google · Official | 50 · Use with care | ★ 21k | 11d ago | Apache-2.0 |
| Google Cloud Data Lineage Summary ✓ Google · Official | 49 · Use with care | ★ 21k | 11d ago | Apache-2.0 |
How did FollowSkills review this skill?
The documentation requires user mode selection and confirmation at important stages, and clearly calls for IRB, FDA, biostatistical, and professional review. However, it does not adequately specify privacy handling for clinical inputs, external-service data flows, file-access boundaries, rollback, or dependency security, so points are deducted.
The workflow contains material inconsistencies: SKILL.md refers to both 05-generate-document.md and 05-concatenate-protocol.md, while the README mentions WebSearch although the research subskill prohibits generic web search. requirements.txt, the claimed sample-size script, and comprehensive tests are not supplied; abnormal-input and failure-feedback handling is also thin, so the static score remains low.
The audience and device/drug scenarios are reasonably clear, and the research-only and full-protocol modes are differentiated. However, trigger boundaries, non-fit ranges, and environment requirements are incomplete; the workflow is strongly centered on U.S. FDA/NIH and Claude Desktop MCP, with no Chinese-language or mainland-China reachability guidance, so points are deducted.
The material is layered across the main skill, README, references, assets, and waypoints, with usage examples and substantial disclaimers. License metadata, versioning, changelog, maintenance ownership, and update path are absent, and filenames conflict with parts of the described workflow, so points are deducted.
The design plausibly covers initialization, regulatory research, modular protocol drafting, statistical calculation, and final assembly, with potentially usable artifacts. Actual completion depends on MCP services, Python dependencies, and files not established by the supplied evidence, and professional review remains necessary; static material cannot verify completeness or comparative benefit over manual work, so the score is capped low.
The supplied files include detailed workflow instructions, subskills, a template, and traceable FDA/regulatory links. However, no test suite, CI workflow, execution record, or third-party reproduction evidence is provided, so the key paths cannot be statically verified.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →