DeepRead Medical Records

Extract structured data from medical records, lab reports, prescriptions, and clinical documents. Pre-built schemas for patient info, diagnoses, medications,...

Security Scan

The skill's requirements and instructions align with its stated purpose (sending documents to a third‑party extraction API using an API key), but it will transmit protected health information to an external service and makes unverified HIPAA/accuracy claims — verify vendor controls before use.

Purpose & Capability

Name/description match the declared runtime behavior: the skill sends clinical documents to an external extraction API and returns structured JSON. The single required env var (DEEPREAD_API_KEY) is appropriate for this purpose.

Instruction Scope

SKILL.md explicitly instructs POSTing documents to https://api.deepread.tech and polling for results — this is expected for an extraction service. Important: that means PHI/PII will be transmitted off‑host. SKILL.md does not instruct reading unrelated system files. It claims PII redaction/HIPAA compliance and 97%+ accuracy; those are vendor claims and not verifiable from the instructions.

Install Mechanism

Instruction‑only skill with no install spec and no code files — nothing is written to disk by an installer. Lowest installation risk.

Credentials

Only one credential is required (DEEPREAD_API_KEY), which is proportional to a remote API integration. Note: the example Python snippet uses a local API_KEY variable rather than DEEPREAD_API_KEY — be sure the agent uses the declared primaryEnv rather than embedding keys in code.

Persistence & Privilege

always is false and the skill is user‑invocable; it does not request permanent/system‑wide presence or modification of other skills/config. Autonomous invocation is allowed by default but is not combined with other privilege escalations here.

Assessment

This skill appears to do what it says: it will upload documents to api.deepread.tech and return structured data. Before installing or using with real patient records, confirm the vendor's legal/security posture (HIPAA BAA, data retention, encryption at rest/in transit, region), and never hardcode API keys in code or prompts. Test with synthetic data to validate redaction quality, check that the agent actually reads DEEPREAD_API_KEY (not an inlined API_KEY), rotate keys regularly, and ensure contract/permissions are in place before sending PHI to a third party. If you cannot confirm appropriate safeguards (BAA, retention policy, access controls), treat this as a risk to patient privacy and avoid uploading real records.

DeepRead Medical Records Processing

Extract structured data from medical records, lab reports, prescriptions, discharge summaries, and clinical documents. Then redact patient PII for HIPAA-compliant sharing, archiving, or LLM processing.

This skill instructs the agent to POST documents to https://api.deepread.tech and poll for results. No system files are modified.

What You Get Back

Submit a medical record and get structured JSON:

{
  "patient_name": {"value": "Jane Smith", "hil_flag": false, "found_on_page": 1},
  "date_of_birth": {"value": "1990-03-15", "hil_flag": false, "found_on_page": 1},
  "mrn": {"value": "MRN-2026-004521", "hil_flag": false, "found_on_page": 1},
  "visit_date": {"value": "2026-03-28", "hil_flag": false, "found_on_page": 1},
  "provider": {"value": "Dr. Sarah Chen, MD", "hil_flag": false, "found_on_page": 1},
  "diagnoses": {"value": [
    {"code": "J06.9", "description": "Acute upper respiratory infection"},
    {"code": "R50.9", "description": "Fever, unspecified"}
  ], "hil_flag": false, "found_on_page": 2},
  "medications": {"value": [
    {"name": "Amoxicillin", "dosage": "500mg", "frequency": "3x daily", "duration": "10 days"},
    {"name": "Acetaminophen", "dosage": "500mg", "frequency": "as needed", "duration": "PRN"}
  ], "hil_flag": false, "found_on_page": 2},
  "vitals": {"value": {
    "blood_pressure": "128/82",
    "heart_rate": 88,
    "temperature": 101.2,
    "weight": "165 lbs"
  }, "hil_flag": false, "found_on_page": 1},
  "follow_up": {"value": "Return in 2 weeks if symptoms persist", "hil_flag": true, "reason": "Handwritten note, partial OCR"}
}

Fields with hil_flag: true need human review. Everything else is high-confidence.

Setup

Get Your API Key

open "https://www.deepread.tech/dashboard/?utm_source=clawhub"

Save it:

export DEEPREAD_API_KEY="sk_live_your_key_here"

Medical Record Schema

Use this pre-built schema for clinical documents. Customize it for your specific document types:

{
  "type": "object",
  "properties": {
    "patient_name": {"type": "string", "description": "Patient full name"},
    "date_of_birth": {"type": "string", "description": "Patient date of birth (YYYY-MM-DD)"},
    "mrn": {"type": "string", "description": "Medical record number or patient ID"},
    "visit_date": {"type": "string", "description": "Date of visit or service (YYYY-MM-DD)"},
    "provider": {"type": "string", "description": "Treating physician or provider name with credentials"},
    "facility": {"type": "string", "description": "Hospital or clinic name"},
    "diagnoses": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "code": {"type": "string", "description": "ICD-10 diagnosis code if present"},
          "description": {"type": "string", "description": "Diagnosis description"}
        }
      },
      "description": "List of diagnoses or conditions"
    },
    "medications": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "name": {"type": "string", "description": "Medication name"},
          "dosage": {"type": "string", "description": "Dosage amount and unit"},
          "frequency": {"type": "string", "description": "How often taken"},
          "duration": {"type": "string", "description": "Duration of treatment"}
        }
      },
      "description": "Prescribed or current medications"
    },
    "vitals": {
      "type": "object",
      "properties": {
        "blood_pressure": {"type": "string", "description": "Blood pressure reading (systolic/diastolic)"},
        "heart_rate": {"type": "number", "description": "Heart rate in BPM"},
        "temperature": {"type": "number", "description": "Body temperature in Fahrenheit"},
        "weight": {"type": "string", "description": "Patient weight with unit"},
        "height": {"type": "string", "description": "Patient height with unit"}
      },
      "description": "Vital signs recorded during visit"
    },
    "procedures": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "code": {"type": "string", "description": "CPT procedure code if present"},
          "description": {"type": "string", "description": "Procedure description"}
        }
      },
      "description": "Procedures performed"
    },
    "lab_results": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "test_name": {"type": "string", "description": "Lab test name"},
          "value": {"type": "string", "description": "Result value with unit"},
          "reference_range": {"type": "string", "description": "Normal reference range"},
          "flag": {"type": "string", "description": "H (high), L (low), or normal"}
        }
      },
      "description": "Laboratory test results"
    },
    "allergies": {
      "type": "array",
      "items": {"type": "string"},
      "description": "Known allergies"
    },
    "follow_up": {"type": "string", "description": "Follow-up instructions or next appointment"},
    "notes": {"type": "string", "description": "Clinical notes or additional observations"}
  }
}

Extract Data From Medical Records

Python

import requests
import json
import time

API_KEY = "sk_live_YOUR_KEY"
BASE = "https://api.deepread.tech"
headers = {"X-API-Key": API_KEY}

schema = json.dumps({
    "type": "object",
    "properties": {
        "patient_name": {"type": "string", "description": "Patient full name"},
        "date_of_birth": {"type": "string", "description": "Patient DOB (YYYY-MM-DD)"},
        "mrn": {"type": "string", "description": "Medical record number"},
        "provider": {"type": "string", "description": "Treating physician with credentials"},
        "diagnoses": {
            "type": "array",
            "items": {"type": "object", "properties": {
                "code": {"type": "string", "description": "ICD-10 code"},
                "description": {"type": "string"}
            }},
            "description": "Diagnoses"
        },
        "medications": {
            "type": "array",
            "items": {"type": "object", "properties": {
                "name": {"type": "string"},
                "dosage": {"type": "string"},
                "frequency": {"type": "string"}
            }},
            "description": "Medications"
        },
        "lab_results": {
            "type": "array",
            "items": {"type": "object", "properties": {
                "test_name": {"type": "string"},
                "value": {"type": "string"},
                "reference_range": {"type": "string"},
                "flag": {"type": "string", "description": "H, L, or normal"}
            }},
            "description": "Lab results"
        }
    }
})

# Submit medical record
with open("patient_record.pdf", "rb") as f:
    job = requests.post(
        f"{BASE}/v1/process",
        headers=headers,
        files={"file": f},
        data={"schema": schema},
    ).json()

job_id = job["id"]
print(f"Processing: {job_id}")

# Poll for results
delay = 5
while True:
    time.sleep(delay)
    result = requests.get(f"{BASE}/v1/jobs/{job_id}", headers=headers).json()

    if result["status"] == "completed":
        data = result["result"]["data"]
        print(json.dumps(data, indent=2))

        # Flag fields needing review
        for field, value in data.items():
            if isinstance(value, dict) and value.get("hil_flag"):
                print(f"\n  REVIEW: {field} — {value.get('reason')}")
        break
    elif result["status"] == "failed":
        print(f"Failed: {result.get('error')}")
        break

    delay = min(delay * 1.5, 15)

cURL

curl -s -X POST https://api.deepread.tech/v1/process \
  -H "X-API-Key: $DEEPREAD_API_KEY" \
  -F "file=@patient_record.pdf" \
  -F 'schema={"type":"object","properties":{"patient_name":{"type":"string","description":"Patient full name"},"mrn":{"type":"string","description":"Medical record number"},"diagnoses":{"type":"array","items":{"type":"object","properties":{"code":{"type":"string","description":"ICD-10 code"},"description":{"type":"string"}}},"description":"Diagnoses"},"medications":{"type":"array","items":{"type":"object","properties":{"name":{"type":"string"},"dosage":{"type":"string"},"frequency":{"type":"string"}}},"description":"Medications"}}}'

HIPAA Compliance: Redact Patient PII

After extracting data, redact patient PII from the original document before sharing, archiving, or sending to an LLM:

# Step 1: Extract structured data (keep the data you need)
with open("patient_record.pdf", "rb") as f:
    extract_job = requests.post(
        f"{BASE}/v1/process",
        headers=headers,
        files={"file": f},
        data={"schema": schema},
    ).json()

# Step 2: Redact PII from the original (clean copy for filing)
with open("patient_record.pdf", "rb") as f:
    redact_job = requests.post(
        f"{BASE}/v1/pii/redact",
        headers=headers,
        files={"file": f},
    ).json()

# Poll for redaction results
redact_id = redact_job["id"]
delay = 5
while True:
    time.sleep(delay)
    result = requests.get(f"{BASE}/v1/pii/{redact_id}", headers=headers).json()
    if result["status"] == "completed":
        report = result["report"]
        print(f"Redacted {report['total_redactions']} PII instances")
        for pii_type, info in report["pii_detected"].items():
            print(f"  {pii_type}: {info['count']} found")

        # Download redacted file
        pdf = requests.get(result["redacted_file_url"]).content
        with open("patient_record_redacted.pdf", "wb") as f:
            f.write(pdf)
        print("Saved: patient_record_redacted.pdf")
        break
    elif result["status"] == "failed":
        print(f"Failed: {result.get('error')}")
        break
    delay = min(delay * 1.5, 15)

DeepRead's PII redaction is context-aware for medical documents:

• "Dr. Sarah Chen" on letterhead — physician, SKIP
• "Sarah Chen" on intake form — patient, REDACT
• "[e-mail]" — institutional email, SKIP
• "[e-mail]" — personal email, REDACT
• Invoice date — not a DOB, SKIP
• "Date of Birth: 03/15/1990" — DOB, REDACT

14 PII types detected: names, SSNs, DOB, medical record numbers, phone, email, address, credit cards, passport, driver's license, bank accounts, IBANs, IPs, URLs.

Use Cases

• Clinical Data Extraction — Pull diagnoses, medications, vitals, and lab results from patient records into your EHR or data warehouse
• Lab Report Processing — Extract test names, values, reference ranges, and abnormal flags from lab reports
• Prescription Digitization — Convert handwritten or printed prescriptions into structured medication data
• Insurance Claims — Extract procedure codes, diagnosis codes, and patient info for claims processing
• Medical Research — De-identify patient records for research datasets while preserving clinical data
• Discharge Summary Processing — Extract follow-up instructions, medications, and diagnoses from discharge paperwork
• Prior Authorization — Pull relevant clinical data to auto-fill prior auth forms

Tips for Medical Documents

• Use ICD-10/CPT code descriptions — Adding "ICD-10 diagnosis code if present" in the schema description helps the model recognize code formats
• Specify units — "Temperature in Fahrenheit" or "Weight with unit (lbs or kg)" reduces ambiguity
• Create blueprints for recurring form types — If you process the same lab report format repeatedly, train a blueprint at deepread.tech/dashboard/optimizer for 20-30% improvement
• Always redact before sharing — Use the PII redaction endpoint before sending documents to LLMs, external systems, or non-clinical staff

BYOK — Zero Processing Costs

Connect your own OpenAI, Google, or OpenRouter key via the dashboard. All document processing routes through your provider — zero DeepRead LLM costs, page quota skipped.

Set it up: https://www.deepread.tech/dashboard/byok

Related DeepRead Skills

• deepread-ocr — General OCR and structured extraction — clawhub install uday390/deepread-ocr
• deepread-pii — Redact PII from any document — clawhub install uday390/deepread-pii
• deepread-form-fill — Fill PDF forms with AI vision — clawhub install uday390/deepread-form-fill
• deepread-invoice — Invoice and receipt processing — clawhub install uday390/deepread-invoice
• deepread-agent-setup — OAuth device flow authentication — clawhub install uday390/deepread-agent-setup
• deepread-byok — Bring Your Own Key setup — clawhub install uday390/deepread-byok

Support

• Dashboard: https://www.deepread.tech/dashboard
• Demo Repo: https://github.com/deepread-tech/deepread-demo
• Issues: https://github.com/deepread-tech/deep-read-service/issues
• Email: [e-mail]

---

Get started free: https://www.deepread.tech/dashboard/?utm_source=clawhub