Implement OpenEvidence reference architecture with best-practice project layout. Use when designing new clinical AI integrations, reviewing project structure, or establishing architecture standards...
Map people, product surfaces, evidence review, governed records, and incident paths without fabricating integrations. Keep inputs minimal, separate observed facts from assumptions, and leave consequential decisions with the named accountable owner.
Use Read, Glob, and Grep to inspect supplied policies, plans, and evidence. Use WebFetch only for current first-party OpenEvidence documentation. Use Write or Edit only when the user requests a named deliverable with an approved destination. Never expose credentials, PHI, recordings, or unrestricted environment output.
Use only the official OpenEvidence web/mobile sign-in or an institution-approved access path. Do not invent API keys, OAuth clients, SDK credentials, service accounts, or private endpoints. Never ask a user to reveal a password, session token, cookie, or recovery code.
Do not create or share accounts; change access, roles, agreements, consent, retention, or security settings; enter PHI; record a conversation; copy content into another system; contact a patient; make a diagnosis or treatment decision; submit billing; transmit a support packet; run a production pilot; or represent vendor capabilities without explicit approval from the accountable owner. A qualified professional remains responsible for clinical decisions.
Return scope, current first-party evidence and date, data classification, workflow or findings, citations reviewed, assumptions rejected, clinical and governance owners, approval state, unresolved risk, and the exact next action. Redact patient and credential data.
| Condition | Response |
|---|---|
| Private endpoint appears in design | Remove it until a signed/current contract documents it. |
| AI output becomes system of record automatically | Insert human review and approved record controls. |
| No downtime path | Block go-live until one exists. |
This compact example shows the minimum reviewable handoff; adapt fields to the approved workflow without adding sensitive data.
Input:
workflow=Ask-to-clinical-note; systems=OpenEvidence+EHR; integration=manual-copy
Expected handoff:
boundaries=6; human-gates=2; undocumented-integrations=0; review=pending