Trial Eligibility AgentSAFE

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

{
"trial_id": "NCT00000000",
"criteria": [
{"id": "I-01", "text": "Age >= 18", "status": "MET", "evidence": "Age 58", "confidence": "high"}
],
"eligibility_summary": "potentially_eligible",
"data_gaps": ["Latest ECOG score"],
"alerts": ["Confirm brain MRI within 30 days"]
}

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: trial-eligibility-agent
description: Parse trial protocols and patient data to produce criterion-level MET/NOT/UNKNOWN determinations with evidence and gaps for clinical trial screening tasks.
allowed-tools:
  - read_file
  - run_shell_command
---

## At-a-Glance
- **description (10-20 chars):** Trial triage hub
- **keywords:** eligibility, ClinicalTrials, FHIR, evidence, gaps
- **measurable_outcome:** Produce a MET/NOT/UNKNOWN matrix with supporting citations for ≥90% of inclusion/exclusion criteria within 5 minutes per trial request.

## Inputs
- `trial_id` (NCT or sponsor ID) plus protocol text if not public.
- `patient_summary` narrative and optional `patient_structured` FHIR bundle.
- Declare data sources used (notes, labs, imaging, meds) to show provenance.

## Outputs
1. Structured table (JSON recommended) listing each criterion id/text with status, evidence snippet, and confidence.
2. Overall recommendation (`potentially_eligible`, `not_eligible`, `needs_more_information`).
3. Data gap checklist covering missing labs/imaging/biomarkers.

## Workflow
1. **Acquire protocol:** Pull eligibility text from ClinicalTrials.gov or sponsor PDF.
2. **Normalize criteria:** Break into atomic checks with AND/OR logic and thresholds.
3. **Extract patient facts:** Map narrative + FHIR data into canonical features (age, labs, ECOG, biomarkers).
4. **Evaluate:** Assign MET/NOT/UNKNOWN with cited evidence for each criterion, flag missing context explicitly.
5. **Summarize:** Present recommendation and highlight gating unknowns plus next-best actions.

## Guardrails
- Never claim enrollment decisions; mark outputs as advisory.
- Cite direct patient evidence for every MET/NOT call; default to UNKNOWN rather than guessing.
- Respect PHI handling expectations—avoid storing raw notes outside secure paths.

## Tooling & References
- Use `README.md` for API snippets (FHIR parsing, JSON schema) and dependency versions.
- Pair with `Clinical/Trial_Matching/TrialGPT` when retrieval/ranking is also needed.


<!-- AUTHOR_SIGNATURE: 9a7f3c2e-MD-BABU-MIA-2026-MSSM-SECURE -->

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.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (1)

Gates applied: no_behavioural_pass.