Trialgpt MatchingSAFE
The largest open-source medical AI skills library for OpenClaw🦞.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Source: ncbi-nlp/TrialGPT Local Repository: ./repo Status: Integrated & Downloaded
Overview
TrialGPT is an NIH-developed framework for matching patients to clinical trials using LLMs. It provides a structured pipeline for trial retrieval, eligibility parsing, and ranking with evidence-based explanations.
Capabilities
- Trial Retrieval - Identify candidate trials from ClinicalTrials.gov.
- Criteria Parsing - Convert eligibility text into structured criteria.
- Patient Profiling - Summarize patient records into matchable features.
- Ranking + Explanation - Score trial relevance and provide justifications.
Recommended Usage
- Install dependencies
cd repo pip install -r requirements.txt
- Run retrieval - identify candidate trials for a condition.
- Run matching - evaluate eligibility with structured criteria.
- Review outputs - validate by clinician or trial coordinator.
Input and Output Expectations
Input:
- Patient summary (structured or narrative)
- Condition keywords or diagnosis codes
Output:
- Ranked trials with relevance scores
- Criteria-level match explanations
- Missing data checklist
Integration Notes
- Use TrialGPT for retrieval and initial ranking, then hand off to the Clinical Trial Eligibility Agent for deeper criterion-by-criterion analysis.
- Cache trial metadata (NCT ID, protocol version) to ensure reproducibility.
Limitations
- Requires up-to-date trial metadata; outdated data can miscl
29f31a89230cOBSERVED · 2026-10-08Install
Commands as the repository documents them. They are shown, not run.
pip install -r requirements.txt
pip install -r requirements.txt
What it tells the agent
The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.
<!-- # COPYRIGHT NOTICE # This file is part of the "Universal Biomedical Skills" project. # Copyright (c) 2026 MD BABU MIA, PhD <[email protected]> # All Rights Reserved. # # This code is proprietary and confidential. # Unauthorized copying of this file, via any medium is strictly prohibited. # # Provenance: Authenticated by MD BABU MIA --> --- name: trialgpt-matching description: Trial shortlist keywords: - retrieval - ranking - ClinicalTrials - patient-profile measurable_outcome: Produce ≥5 ranked trials (when available) with rationale + missing-data notes within 3 minutes of receiving a patient query. license: MIT metadata: author: TrialGPT Team version: "1.0.0" compatibility: - system: Python 3.9+ allowed-tools: - run_shell_command - read_file --- # TrialGPT Matching Run the locally checked-out TrialGPT pipeline to retrieve, rank, and explain candidate trials for a patient before deeper eligibility review. ## Inputs - Patient summary (structured JSON or free text) with condition keywords. - Optional filters: geography, phase, intervention, biomarker. - Up-to-date ClinicalTrials.gov dump or API access. ## Outputs - Ranked trial table with NCT ID, title, score, and short justification. - Parsed inclusion/exclusion text ready for downstream eligibility agents. - Missing data checklist (e.g., "ECOG not provided"). ## Workflow 1. **Setup:** `cd repo && pip install -r requirements.txt` (or reuse env). 2. **Trial retrieval:** Run TrialGPT retriever to pull candidate trials for the indication. 3. **Criteria parsing:** Convert eligibility blocks to structured criteria JSON. 4. **Patient profiling:** Summarize patient facts (labs, prior therapies, biomarkers). 5. **Ranking:** Execute TrialGPT ranking script to score each trial and emit explanations. 6. **Handoff:** Export ranked list + structured criteria for `trial-eligibility-agent`. ## Guardrails - Refresh ClinicalTrials.gov metadata regularly to avoid stale trials. - Label scores as AI-generated suggestions pending clinician validation. - Retain prompt/config metadata for audit trails. ## References - Detailed usage instructions and repo layout live in `README.md`. - Coordinate with `Skills/Clinical/Trial_Eligibility_Agent` for criterion-level review. <!-- AUTHOR_SIGNATURE: 9a7f3c2e-MD-BABU-MIA-2026-MSSM-SECURE -->
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- none-observed
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (5)
repo/dataset/sigir/corpus.jsonl
repo/dataset/sigir/retrieved_trials.json
repo/dataset/trec_2021/retrieved_trials.json
repo/dataset/trec_2022/retrieved_trials.json
Gates applied: no_behavioural_pass.
29f31a89230cfull audit observations/trust-audit/skill/freedomintelligence__trialgpt-matching.json · Report an issue / request a re-scanAudit history
Every audit this skill has had.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-08 | 29f31a89230c | SAFE | B | 89 | first audit |
Questions
What does the Trialgpt Matching skill do?
The largest open-source medical AI skills library for OpenClaw🦞.
Is Trialgpt Matching safe to install?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean skill reads B.
What can Trialgpt Matching access on my machine?
The audit observed no filesystem, network or shell use at all in its source.
How current is this page?
The grade is for one exact copy of the source (29f31a89230c), read on 2026-10-08. The repository is watched, and a new audit runs when it changes — this is the first audit.