Agentic workbench

The agent drafts. The laboratory director still signs.

Validation packets, SOP skeletons, and change synthesis with citations and an audit trail. In the product, every draft is stored, queued, and accepted or rejected by a second human. Nothing here becomes an operational commitment until that gate fires.

Validation / verification outline

A CLIA-shaped skeleton for accuracy, precision, reportable range, reference interval, and analytic sensitivity. Human review is mandatory before the packet is real.

SOP and competency outline

Principle, specimen, reagents, procedure, QC, reporting, limitations, and a competency checklist keyed to the method, not a generic HR form.

Jurisdictional mapping

Federal baseline plus a placeholder for the state overlay: reportable condition, isolate submission, and notification timing.

What changed this week

A synthesis of the living change log into suggested actions. Suggestions are not instructions. A director still decides.

Human kill-switch

Every persona proposal lands in a queue. Approve writes an audit event. Reject fires the kill switch. Nothing accessions, verifies, or transmits without that gate.

Try it

A CLIA-shaped outline from the living catalog.

Template generation, not a model talking off-script. Pick a test. Download the draft. A person still completes the science.

Lands in VIRVirology. Last reviewed 2026-08-13.

# CLIA-aligned validation / verification outline

Test: SARS-CoV-2 RT-PCR (COV2)
Method: NAA with probe detection
Owning lab: Virology (VIR)
Pathogen: SARS-CoV-2
Reviewed against catalog: 2026-08-13

Status: DRAFT - human review required. This is decision support, not a completed validation.

## 1. Intended use
- Specimen types: Nasopharyngeal swab, saliva, lower respiratory
- Claim: Still the backbone of respiratory situational awareness. Variant context now lives in surveillance, not in every diagnostic report.
- What this assay does not claim: patient-level diagnosis without medical director authority.

## 2. Regulatory posture
- High complexity. Verification of an FDA-authorized assay, or full validation if run as an LDT.
- Post-2025 LDT status: LDTs remain under CLIA. Flexibility requires documented validation.

## 3. Performance characteristics to demonstrate
- Accuracy / comparison to a reference or predicate method
- Precision (repeatability and reproducibility)
- Analytic sensitivity (LoD) and specificity / cross-reactivity
- Reportable range (if quantitative or semi-quantitative)
- Specimen-type equivalence for every matrix you will report
- Interfering substances relevant to Nasopharyngeal swab, saliva, lower respiratory

## 4. Result values (coded first)
- SARS-CoV-2 RNA (LOINC 94500-6)
  - Detected (SNOMED 260373001)
  - Not detected (SNOMED 260415000)
  - Inconclusive (SNOMED 419984006)

## 5. Quality control and PT
- External positive and negative controls each run. Proficiency testing per CLIA. Lot-to-lot checks on new reagent lots.

## 6. Reflex, referral, reporting
- Reflex: Unexpected positives in low-prevalence settings may reflex to a second target or a reference method.
- Reportability: Reportable in most jurisdictions. Outbreak clusters escalate to epidemiology the same day.
- ELR: OBX-3 LOINC 94500-6. OBX-5 SNOMED 260373001 / 260415000. Do not put a panel LOINC in every OBX-3.

## 7. Human gates
- Medical director (or designee) signs the validation summary.
- Quality manager files the packet and assigns an SOP version.
- No agent or template may mark this assay live.

## 8. Sources to attach
- CDC LIVD SARS-CoV-2 mappings; US Realm ELR IG.
- Primary package insert or LDT design record
- Raw data tables (not only the narrative)