Workflow chains
AuraScore 81/100

Clinical Trial Protocol Eligibility Agent Workflow Blueprint

Design a multi-stage agentic workflow chain for clinical trial matching, EHR screening, and investigator verification.

Use this template when architecting an automated multi-step clinical trial patient matching pipeline across hospital data systems. It is ideal for clinical informatics leaders seeking to balance automated EHR parsing with strict regulatory validation.

Template

Role: Principal Clinical Informatics Architect specializing in regulatory-grade agentic orchestration systems.

Context

  • Target Protocol: {{trial_protocol_id}}
  • Primary EHR Architecture: {{ehr_source_system}}
  • Clinical Target Condition: {{target_condition}}
  • Regulatory Compliance Standard: {{regulatory_framework}}
  • Human Oversight Policy: {{human_review_policy}}
  • Target Processing Volume: {{throughput_target}}

Task

Develop a comprehensive architectural report outlining a deterministic multi-agent workflow chain that ingests semi-structured EHR records, evaluates inclusion and exclusion criteria against {{trial_protocol_id}}, and routes verified candidates to clinical coordinators.

Method

  1. Map the end-to-end agentic workflow topology from raw record ingestion in {{ehr_source_system}} to candidate presentation.
  2. Define the schema validation agent responsible for sanitizing and normalizing {{target_condition}} pathology, lab, and demographic data.
  3. Detail the deterministic logic agent for structured inclusion/exclusion criteria parsing with zero-hallucination guardrails.
  4. Formulate the unstructured extraction agent protocol for parsing oncology and physician notes against {{regulatory_framework}} standards.
  5. Design the reconciliation and confidence-scoring agent that calculates protocol fit scores under {{human_review_policy}}.
  6. Architect the state persistence and rollback mechanism to maintain session integrity during high-volume processing of {{throughput_target}}.
  7. Specify human-in-the-loop (HITL) escalation triggers for ambiguous clinical annotations and borderline biomarker readings.
  8. Establish the audit logging and traceability mechanism for all intermediate agent reasoning steps.

Constraints

  • MUST enforce {{human_review_policy}} as a blocking gate before any prospective patient identity is exposed to researchers.
  • MUST NOT allow non-deterministic agent outputs to overrule hard laboratory exclusion limits.
  • All data transformations MUST comply with {{regulatory_framework}} audit trail requirements.
  • The workflow must scale to handle {{throughput_target}} without state corruption.

Output format

Provide a technical report with these exact section headings:

  1. Executive Workflow Architecture
  2. Agent Chain Sequence & Data Contracts
  3. Exception Handling & Confidence Gating
  4. Regulatory Traceability & Audit Schema
  5. Infrastructure & Throughput Feasibility Total word count must be between 900 and 1,300 words.

Self-review

  • Are all 6 variables integrated contextually into the pipeline logic?
  • Does the report clearly define state management between individual agent nodes?
  • Is the human-in-the-loop escalation mechanism strictly non-bypassable?
AuraScore breakdown
81/100Provisional
Instruction clarity15/15 · Strong

Explicit role, a named task, and discrete steps the model can follow.

Context architecture12/12 · Strong

Background, inputs and variables the model needs before it starts.

Constraint engineering12/12 · Strong

Hard boundaries — what the model must and must not do.

Output specification6/14 · Thin

A named, field-level shape for the response.

Reasoning structure10/10 · Strong

Ordered work items that force analysis before an answer.

Model compatibility10/10 · Strong

Length and structure that travel across frontier models.

Token efficiency5/10 · Thin

Signal density — instruction weight without padding.

Reusability7/7 · Strong

Documented variables so the scaffold adapts to new inputs.

Robustness3/5 · Adequate

Quality bar, assumptions and behaviour when inputs are thin.

Observed performance1/5 · Thin

How much real usage the template has behind it.

ai-agents
agents-workflows
healthcare-life-sciences
clinical-trials
agent-chains
health-informatics