SCRIMED turns every limitation into a safe operating path, owner, proof route, and graduation gate.

Can we evaluate SCRIMED without exposing PHI or live operations?

• Proof route: /health-records

Will SCRIMED overpromise compliance, security, or clinical readiness?

• Proof route: /trust-center

What happens when a buyer asks for something outside today’s capability?

• Proof route: /limitations-workarounds

How does SCRIMED keep reliability issues from becoming hidden risk?

• Proof route: /trust-safety-operations

Buyer asks to upload, paste, connect, or analyze PHI, patient identifiers, member IDs, live charts, production records, or customer credentials.

• Safe response: SCRIMED can evaluate the workflow with synthetic fixtures, metadata-only source references, and a health-record safety plan while PHI authority remains gated.
• Path: Health Records Safety Exchange, Clinical Authority Readiness, Boundary Resolution, Qualified customer/privacy approval
• Hard stops: PHI introduced, patient identifier introduced, production credential requested, live chart requested
• Evidence: BAA/DPA when required, customer authority, security review, clinical governance approval, connector/data boundary approval
• Proof routes: /health-records, /clinical-authority-readiness, /boundary-resolution

Buyer, demo, sales, or product language asks SCRIMED to diagnose, treat, triage, prescribe, route patients, sign notes, or provide production CDS.

• Safe response: SCRIMED can prepare workflow intelligence, draft queues, evidence organization, and governance packets, but live clinical decisions require qualified authority.
• Path: Clinical Care Activation, Clinical Authority Readiness, QA Claim Guard, Regulatory/intended-use review
• Hard stops: diagnosis implied, treatment implied, autonomous triage requested, clinician signature implied
• Evidence: intended-use review, clinical governance approval, customer scope, monitoring/override plan, rollback evidence
• Proof routes: /clinical-care-activation, /clinical-authority-readiness, /qa-claim-guard

Buyer requests production EHR, HIE, payer, imaging, device, SMART launch, writeback, patient matching, or payer submission connectivity.

• Safe response: SCRIMED can provide standards-aware synthetic conformance review, connector contract planning, and no-mutation evidence before production connectivity is approved.
• Path: Interoperability registry, Health Records Safety Exchange, Platform Power API contract register, Customer sandbox authority
• Hard stops: production endpoint requested, EHR writeback requested, payer submission requested, patient matching requested
• Evidence: customer sandbox, security review, connector contract, mutation policy, audit/monitoring, rollback plan
• Proof routes: /interoperability, /health-records, /platform-power, /integrations/fixture-validation

Procurement asks for SOC 2, HITRUST, ISO, HIPAA certification, penetration testing, vendor-risk approval, BAA/DPA, or security signoff.

• Safe response: SCRIMED can provide trust-center readiness evidence, metadata-only artifact references, and a review plan without claiming certification or approval.
• Path: Trust Center, Global Certification Readiness, Provider Security Reviews, Procurement Evidence Registry
• Hard stops: security certified claimed, SOC 2 certified claimed, HITRUST certified claimed, penetration test complete claimed
• Evidence: qualified assessment, remediation evidence, approved artifact reference, customer-specific acceptance, release authority
• Proof routes: /trust-center, /global-certification-readiness, /pilot-workspace/access

Buyer asks for public API access, unlimited usage, public SDK, uptime, support coverage, contractual SLA, managed service coverage, or trillion-scale capacity.

• Safe response: SCRIMED can share route summaries, version posture, boundary headers, rate-class planning, and support assumptions without creating an SLA.
• Path: Platform Power, Enterprise Scalability, Service Reliability, Enterprise Business Ops
• Hard stops: public API SLA implied, unlimited usage promised, contractual uptime promised, managed service commitment implied
• Evidence: contract terms, rate limits, support tier, incident response plan, monitoring, price model, executive approval
• Proof routes: /platform-power, /enterprise-scalability, /service-reliability, /enterprise-business-ops

User asks agents to send messages, create invites, execute tools, remediate production, make clinical/legal/financial decisions, or act without human approval.

• Safe response: SCRIMED agents can plan, inspect, draft, evaluate, and route work while protected execution remains human-approved and audit-bound.
• Path: AgentOS, TrustOS, Continuous Review, Manual QA Execution Console, QA Claim Guard
• Hard stops: tool execution without approval, production remediation requested, clinical decision requested, legal or financial decision requested
• Evidence: allowed tool schema, blocked tool list, human approval UI, audit persistence, rollback behavior, customer authority
• Proof routes: /agents, /trust-os, /continuous-review-audit, /qa-manual-execution-console, /qa-claim-guard

Proposal, investor, buyer, or board language asks for legal advice, tax/accounting conclusions, audited financials, valuation, securities material, ROI, revenue, reimbursement, or profit guarantees.

• Safe response: SCRIMED can provide readiness evidence, fixed-scope offers, buyer-approved measurement plans, and review slots without giving professional advice or guarantees.
• Path: Enterprise Business Ops, Growth Engine, Capital Vitality, Investor Audience Readiness, QA Claim Guard
• Hard stops: legal advice implied, tax advice implied, securities material implied, ROI or revenue guaranteed
• Evidence: qualified review, approved language, buyer baseline, measurement plan, recipient context, release decision
• Proof routes: /enterprise-business-ops, /growth-engine, /capital-vitality, /investor-audience-readiness, /qa-claim-guard

Buyer or partner asks for country launch, public-sector approval, government endorsement, GDPR/EU AI Act conformity, NHS/MHRA/Australia approval, or data-residency approval.

• Safe response: SCRIMED can prepare no-PHI regional packs, deployment questions, and evidence implications without claiming local approval or conformity.
• Path: Global Reach, Global Certification Readiness, Deployment Profiles, Boundary Resolution
• Hard stops: country launch approved claimed, government endorsement implied, data residency approved claimed, regional compliance approved claimed
• Evidence: regional counsel review, privacy/security review, hosting decision, procurement authority, partner/customer signoff
• Proof routes: /global-reach, /global-certification-readiness, /deployment-profiles, /boundary-resolution

live PHI, patient identifiers, member data, live charts, and production credentials

• Signals: PHI upload, patient identifier, member ID, live chart, production credential
• Safe alternative: Use the Synthetic no-PHI packet, Health Records Safety Exchange source mapping, and external evidence references without copying protected records into SCRIMED.
• Approvals: privacy review, security review, customer authority, BAA/DPA when required, clinical governance review
• Validate: npm run smoke:limitations-workarounds
• Fail closed: Requests that contain PHI or identifiers remain blocked before storage, model routing, connector execution, or buyer-facing output.
• Residual risk: Synthetic proof can support diligence, but it does not validate live PHI operations or patient-specific safety.
• Hard stops: PHI pasted into prompt, live chart uploaded, patient matching requested, production credential provided
• Proof routes: /health-records, /clinical-authority-readiness, /boundary-resolution, /limitations-workarounds

diagnosis, treatment, prescribing, triage, patient routing, signed notes, and production CDS

• Signals: diagnosis, treatment recommendation, prescribing, triage, sign note, clinical decision support
• Safe alternative: Use clinical-governance preparation, synthetic clinical robustness evaluation, evidence organization, and human review queues.
• Approvals: qualified clinician review, intended-use review, regulatory classification, customer scope approval, monitoring and override plan
• Validate: npm run smoke:clinical-robustness-lab
• Fail closed: Clinical outputs remain research/demo or draft-only and cannot become autonomous care, diagnosis, treatment, prescribing, or signed documentation.
• Residual risk: Reviewer-gated drafts can reduce operational burden but do not replace clinician judgment or establish clinical validation.
• Hard stops: autonomous diagnosis, treatment plan finalized, prescription action, patient routed without approval
• Proof routes: /clinical-care-activation, /clinical-authority-readiness, /clinical-robustness-lab, /qa-claim-guard

production EHR/HIE/payer/imaging/device connector use, patient matching, writeback, and mutation

• Signals: production endpoint, SMART launch, EHR writeback, patient matching, PACS mutation, device feed
• Safe alternative: Use fixture validation, interoperability conformance metadata, connector contract review, and no-mutation evidence packets.
• Approvals: customer sandbox authority, security review, connector contract, mutation policy, audit and rollback review
• Validate: npm run smoke:public
• Fail closed: Production connector, writeback, patient matching, and raw connector payload logging stay blocked until explicit customer and technical authority exists.
• Residual risk: Synthetic conformance shows readiness posture only; it cannot prove customer-specific connector acceptance.
• Hard stops: production endpoint requested, raw connector payload stored, EHR writeback requested, patient merge requested
• Proof routes: /interoperability, /health-records, /integrations/fixture-validation, /platform-power

payer submission, billing submission, claim filing, reimbursement certainty, and medical-necessity determination

• Signals: submit prior authorization, submit claim, bill payer, medical necessity approved, reimbursement guaranteed
• Safe alternative: Use Documentation-Before-Authorization checks, RCM denial-risk metadata, payer-policy lookup scaffolds, and human-reviewed packets.
• Approvals: payer workflow owner, clinical reviewer, billing compliance review, customer authorization, release decision
• Validate: npm run smoke:documentation-before-authorization
• Fail closed: SCRIMED may identify missing documentation and draft review packets, but it does not submit payer transactions or guarantee reimbursement.
• Residual risk: Documentation readiness can lower denial risk but cannot assure payer behavior or reimbursement.
• Hard stops: payer transaction submitted, claim filed, reimbursement guaranteed, medical necessity certified
• Proof routes: /scrimed-build-roadmap, /enterprise-business-ops, /qa-claim-guard, /limitations-workarounds

autonomous agent execution, production remediation, model routing, tool calls, email/calendar actions, and protected workflow actions

• Signals: execute tool, send email, create calendar invite, auto-remediate, route PHI to model, production model approved
• Safe alternative: Use the Meta-Harness, model-route register, manual QA execution console, and no-secret AAL2 operator lanes.
• Approvals: tool permission owner, TrustOS review, security review, cost owner, clinical/legal/finance owner when relevant
• Validate: npm run smoke:scrimed-meta-harness
• Fail closed: Agents can plan, inspect, draft, evaluate, and recommend, but protected execution requires permissioned tools, audit logs, and human approval.
• Residual risk: Human-reviewed recommendations still require monitoring for drift, misuse, cost spikes, and unsafe automation pressure.
• Hard stops: tool execution without approval, patient outreach, production remediation, clinical/legal/financial decision
• Proof routes: /agents, /trust-os, /scrimed-build-roadmap, /continuous-review-audit, /qa-manual-execution-console

SOC 2, HITRUST, ISO, HIPAA certification, penetration test completion, BAA/DPA, and security signoff

• Signals: SOC 2 certified, HITRUST certified, HIPAA certified, pentest complete, BAA approved, vendor risk approved
• Safe alternative: Use Trust Center readiness, global certification readiness, provider security review metadata, and external evidence references.
• Approvals: qualified assessor, security owner, privacy owner, legal owner, customer-specific acceptance
• Validate: npm run smoke:limitations-workarounds
• Fail closed: Security and privacy claims remain readiness-only unless qualified evidence and approved release authority exist.
• Residual risk: Readiness artifacts can support procurement but do not substitute for independent assessment or customer acceptance.
• Hard stops: certification claimed, assessment implied complete, security signoff promised, BAA accepted without review
• Proof routes: /trust-center, /global-certification-readiness, /pilot-workspace/access, /boundary-resolution

public API SLA, unlimited usage, contractual uptime, managed service coverage, support guarantee, and scale-equivalence claims

• Signals: public API, unlimited usage, uptime guaranteed, managed service, support guarantee, trillion-dollar scale
• Safe alternative: Share route summaries, version posture, boundary headers, draft quotas, rate classes, support assumptions, and no-SLA language.
• Approvals: platform owner, service reliability owner, legal, finance, executive approver
• Validate: npm run smoke:public
• Fail closed: Public routes remain inspectable readiness surfaces and do not create contractual SLA, unlimited scale, or managed-service obligations.
• Residual risk: Technical availability checks are not contractual commitments and do not prove enterprise support capacity.
• Hard stops: unlimited usage promised, contractual uptime promised, managed service active claimed, scale parity guaranteed
• Proof routes: /platform-power, /enterprise-scalability, /service-reliability, /enterprise-business-ops

legal advice, tax/accounting conclusions, audited financials, securities material, valuation, ROI, revenue, reimbursement, and profit guarantees

• Signals: legal advice, tax advice, audited financials, securities material, valuation, ROI guaranteed, revenue guaranteed
• Safe alternative: Use fixed-scope offers, buyer-approved measurement plans, review slots, finance methodology gates, and no-guarantee language.
• Approvals: qualified counsel, finance reviewer, accounting reviewer, tax reviewer when relevant, executive release authority
• Validate: npm run smoke:limitations-workarounds
• Fail closed: Commercial and investor material remains readiness and operating discipline, not legal advice, audited reporting, securities material, or guarantees.
• Residual risk: Value narratives remain estimates or hypotheses until buyer-approved baselines and qualified review exist.
• Hard stops: legal advice provided, tax/accounting conclusion, securities material released, ROI guaranteed
• Proof routes: /enterprise-business-ops, /growth-engine, /capital-vitality, /investor-audience-readiness

country launch, public-sector approval, government endorsement, data residency, GDPR, EU AI Act, NHS, MHRA, and regional procurement claims

• Signals: country launch, government endorsed, data residency approved, GDPR approved, EU AI Act conformant, NHS approved
• Safe alternative: Use global regional workaround packs, deployment profiles, regional evidence implications, and local-counsel owner assignment.
• Approvals: regional counsel, privacy review, security review, hosting decision, procurement authority, partner/customer signoff
• Validate: npm run smoke:limitations-workarounds
• Fail closed: Regional work remains discovery and localization planning, not local approval, procurement acceptance, conformity, or launch authority.
• Residual risk: Global readiness maps requirements but does not resolve local law, procurement, hosting, or clinical authority.
• Hard stops: country launch approved, government endorsement implied, data residency approved early, regional compliance approved
• Proof routes: /global-reach, /global-certification-readiness, /deployment-profiles, /boundary-resolution

customer go-live, buyer release, protected proof distribution, production support, and launch approval

• Signals: go-live approved, buyer release, send proof packet, production support, launch approved
• Safe alternative: Use QA Completion Bridge, Activation Seal, Manual QA Execution Console, Buyer Proof Release, and no-secret protected evidence packets.
• Approvals: approved AAL2 operator, reviewer signoff, release steward, recipient authority, claim guard
• Validate: npm run smoke:aal2:readiness
• Fail closed: Public smoke and synthetic proof do not become buyer release authority or customer go-live approval.
• Residual risk: Protected proof can expire or become stale after deployment, route, reviewer, workspace, recipient, or claim-language changes.
• Hard stops: token retained, buyer release claimed early, go-live approved without signoff, protected proof shared without recipient authority
• Proof routes: /release-continuity, /qa-manual-execution-console, /buyer-release-control-run, /pilot-workspace/access

public quantum capability, quantum clinical advantage, quantum-safe certification, and future infrastructure superiority

• Signals: quantum capability, quantum advantage, quantum-safe certified, future infrastructure superiority
• Safe alternative: Use internal research backlog items with hypothesis, source log, review owner, and no-public-claim labels.
• Approvals: internal research owner, security review, legal review, claim guard, release decision
• Validate: npm run smoke:limitations-workarounds
• Fail closed: Quantum remains internal research and cannot appear as a public capability, certification, clinical advantage, or investor claim.
• Residual risk: Future-facing language can quickly become an unsupported superiority claim if not kept behind release review.
• Hard stops: public quantum capability claimed, quantum clinical advantage claimed, quantum-safe certification claimed
• Proof routes: /continuous-review-audit, /qa-claim-guard, /boundary-resolution, /limitations-workarounds

preflight-live-phi-upload

• Matched playbooks: playbook-live-phi-request, playbook-ehr-writeback-connector-request
• Safe alternative: Use the Synthetic no-PHI packet, Health Records Safety Exchange source mapping, and external evidence references without copying protected records into SCRIMED. Use fixture validation, interoperability conformance metadata, connector contract review, and no-mutation evidence packets.
• Approvals: privacy review, security review, customer authority, BAA/DPA when required, clinical governance review, customer sandbox authority, connector contract, mutation policy, audit and rollback review
• Validate: npm run smoke:limitations-workarounds, npm run smoke:public
• External execution allowed: no
• PHI processing allowed: no
• Autonomous action allowed: no
• Audit: scrimed-limit-6aad07c3

preflight-payer-submit

• Matched playbooks: playbook-payer-submission-request
• Safe alternative: Use Documentation-Before-Authorization checks, RCM denial-risk metadata, payer-policy lookup scaffolds, and human-reviewed packets.
• Approvals: payer workflow owner, clinical reviewer, billing compliance review, customer authorization, release decision
• Validate: npm run smoke:documentation-before-authorization
• External execution allowed: no
• PHI processing allowed: no
• Autonomous action allowed: no
• Audit: scrimed-limit-ddf12661

preflight-agent-remediate

• Matched playbooks: playbook-autonomous-agent-action-request
• Safe alternative: Use the Meta-Harness, model-route register, manual QA execution console, and no-secret AAL2 operator lanes.
• Approvals: tool permission owner, TrustOS review, security review, cost owner, clinical/legal/finance owner when relevant
• Validate: npm run smoke:scrimed-meta-harness
• External execution allowed: no
• PHI processing allowed: no
• Autonomous action allowed: no
• Audit: scrimed-limit-286ca658

preflight-security-certification

• Matched playbooks: playbook-security-certification-request
• Safe alternative: Use Trust Center readiness, global certification readiness, provider security review metadata, and external evidence references.
• Approvals: qualified assessor, security owner, privacy owner, legal owner, customer-specific acceptance
• Validate: npm run smoke:limitations-workarounds
• External execution allowed: no
• PHI processing allowed: no
• Autonomous action allowed: no
• Audit: scrimed-limit-bd76bab8

preflight-safe-synthetic-assessment

• Matched playbooks: none
• Safe alternative: Use a buyer-safe synthetic assessment, metadata-only references, and human-reviewed output language.
• Approvals: human review before external use
• Validate: npm run smoke:limitations-workarounds
• External execution allowed: no
• PHI processing allowed: no
• Autonomous action allowed: no
• Audit: scrimed-limit-a4d43dd8

Future AAL2 durable-store happy-path smoke runs still require an authorized tenant-admin, pilot-lead, or reviewer to supply a fresh short-lived AAL2 bearer token.

• Workaround: Run npm run smoke:aal2:readiness first, then use the no-secret helper with an explicit token source: npm run smoke:aal2:token -- --clipboard-token --clear-clipboard --write-env-local, or the hidden prompt path when clipboard transfer is blocked.
• Resolution: Retain only no-secret smoke status, command names, workspace slug, role class, audit metadata, and packet hashes after a human AAL2 operator runs the strict smoke; never store or paste bearer tokens.
• Next proof: npm run smoke:aal2:durable-store:strict
• Fail-closed check: Missing SCRIMED_BEARER_TOKEN and invalid tokens stop before authenticated mutation; unauthenticated durable-store endpoints remain 401 or 503.
• Gate: Fresh AAL2 session, authorized role, gitignored .env.local or shell env only, strict smoke pass, no-token evidence packet, reviewer signoff, and re-run before buyer-specific proof is released.
• Hard stops: token pasted into source, token retained in docs, AAL2 bypass attempted, observer role used for writes
• Proof routes: /release-continuity, /qa-manual-execution-console, /api/workflows/execution-attempts/durable-store

The deployed app exposes durable-store contracts while protected writes depend on the SCRIMED_EXECUTION_ATTEMPT_DURABLE_STORE_ENABLED runtime flag and protected Supabase RPCs.

• Workaround: Keep public summary and unauthenticated fail-closed smoke active; use strict AAL2 smoke only with a short-lived authorized operator token and synthetic workspace.
• Resolution: Keep the flag paired with named synthetic canaries, migration evidence, runtime-token configuration, rollback owner, strict AAL2 proof, and protected audit review.
• Next proof: npm run smoke:aal2:durable-store
• Fail-closed check: When protected writes are disabled or auth context is missing, authenticated and unauthenticated mutation attempts do not create durable records.
• Gate: Applied migration, server runtime token, feature flag scoped to target, strict AAL2 canary pass, rollback plan, and retained audit event review.
• Hard stops: feature flag enabled without migration, production write enabled without AAL2 smoke, rollback owner missing
• Proof routes: /api/workflows/execution-attempts/durable-store, /workflows/execution-attempts, /release-continuity

The local sandbox may fail DNS resolution for app.scrimedsolutions.com, which can block live production smoke checks inside the default sandbox.

• Workaround: Retry the exact same smoke command with approved network escalation, or run against a local dev server with SCRIMED_BASE_URL=http://127.0.0.1:3025.
• Resolution: Keep every live-domain smoke command reproducible, documented, and paired with a local-server equivalent where practical.
• Next proof: SCRIMED_BASE_URL=https://app.scrimedsolutions.com npm run smoke:public
• Fail-closed check: DNS failure is classified as an environment boundary, not an application pass; the same smoke must pass with network access before release proof expands.
• Gate: Live smoke passes from an approved network path and local fallback remains documented for sandbox-limited development.
• Hard stops: treating ENOTFOUND as product proof, changing application code to bypass DNS, skipping live-domain smoke before release
• Proof routes: /launch-readiness, /release-continuity, /service-reliability

Supabase leaked-password protection and password-auth posture must be resolved before password sign-in is used for protected operations.

• Workaround: Keep protected durable-store and QA flows on short-lived AAL2 sessions from passkey or magic-link paths; exclude password sign-in from protected smoke instructions.
• Resolution: Clear the Supabase security advisor item, document the identity posture, and rerun protected AAL2 smoke before broadening password-based operator access.
• Next proof: npm run smoke:aal2:token -- --prompt-token
• Fail-closed check: Protected APIs continue to require verified Supabase Auth, aal=aal2, session_id, tenant role, and route-level authorization.
• Gate: Advisor clean or password auth excluded, MFA enrollment verified, AAL2 claims present, and tenant role proof retained without secrets.
• Hard stops: claiming HIPAA/SOC certification, using password-only protected access, bypassing AAL2
• Proof routes: /global-certification-readiness, /trust-center, /pilot-workspace/access

Local Next.js builds may warn that the native Darwin SWC binary has a macOS code-signature mismatch and fall back to WASM bindings.

• Workaround: Treat the warning as local toolchain noise when npm run build exits 0; preserve the full build result in release notes.
• Resolution: Refresh local dependencies or reinstall the native SWC package in a clean workspace when the team wants faster native builds.
• Next proof: npm run build
• Fail-closed check: No release proof is accepted unless build exits 0, lint and typecheck pass, and route generation completes.
• Gate: Clean native SWC load or accepted WASM fallback with successful build, typecheck, lint, and smoke evidence.
• Hard stops: ignoring nonzero build exit, hiding build warnings from release notes, treating local toolchain warning as production capacity proof
• Proof routes: /service-reliability, /release-continuity, /platform-power

The SCRIMED worktree can carry many staged, modified, and untracked build artifacts across long execution sessions.

• Workaround: Use git status --short, git diff --name-only, and focused smoke outputs before summarizing; never revert unrelated user or prior-session changes.
• Resolution: Create a release checkpoint with scoped files, generated-output policy, commit message, tag, and deployment evidence after human review.
• Next proof: git status --short
• Fail-closed check: No commit, deploy, or buyer proof claim is made until changed files, untracked files, smoke commands, and known exclusions are listed.
• Gate: Reviewed diff, successful lint/typecheck/build/smoke, clean or intentionally scoped worktree, and release decision recorded.
• Hard stops: reverting unrelated work, committing secrets, claiming release readiness without status and smoke evidence
• Proof routes: /release-continuity, /navigation, /api/product/console

Tenant-admin workspace bootstrap completed for synthetic canary

• Upgrade: Created the protected tenant/workspace membership path for atlas-synthetic-evaluation so authorized AAL2 smoke can prove tenant role, status, and workspace scope before mutation.
• Evidence retained: No-secret membership facts only: workspace slug, active tenant-admin role class, status, and access-review due date.
• Verify: npm run smoke:aal2:token -- --clipboard-token --clear-clipboard --write-env-local
• Rollback/fallback: Suspend membership or downgrade role in Supabase, then rerun protected smoke to confirm record/replay/review return 401 or 403.
• Residual boundary: Future users still need a governed tenant-access workflow, access review, and offboarding path before they can operate protected workflows.
• Next control: Promote the founder bootstrap into a reusable tenant-admin access-review checklist with least-privilege role assignment and expiry evidence.
• Hard stops: role granted without owner, inactive membership used, workspace slug guessed, access review omitted
• Proof routes: /pilot-workspace/access, /release-continuity, /limitations-workarounds

AAL2 token helper now prefers explicit operator token sources

• Upgrade: The helper now prefers explicit session-file, clipboard, or prompt sources before local environment fallbacks and clears the macOS clipboard when requested.
• Evidence retained: No bearer token retained in source, docs, logs, or smoke output; only redacted preflight status and token policy metadata are shown.
• Verify: npm run smoke:aal2:policy-test
• Rollback/fallback: Use the hidden prompt mode and delete SCRIMED_BEARER_TOKEN from .env.local before retrying if clipboard controls are unavailable.
• Residual boundary: A valid JWT still depends on a live AAL2 session and may expire before the protected smoke completes.
• Next control: Add the token-helper preflight to every protected smoke runbook and keep all token-like values behind redaction.
• Hard stops: token printed, token committed, stale token reused, AAL2 bypass attempted
• Proof routes: /release-continuity, /qa-manual-execution-console, /limitations-workarounds

Durable-store PHI guard precision migrations applied

• Upgrade: Applied precision migrations so identifier checks use bounded patterns while preserving hard stops for PHI, patient identifiers, live charts, member data, and production records.
• Evidence retained: Migration file names, contract-check pass state, and synthetic no-PHI smoke result only; no patient data or live records are used.
• Verify: npm run smoke:execution-attempt-durable-store
• Rollback/fallback: Disable protected writes with SCRIMED_EXECUTION_ATTEMPT_DURABLE_STORE_ENABLED=false and keep public fail-closed smoke active while the guard is reviewed.
• Residual boundary: Guard precision is a synthetic-safety control, not permission to process PHI or live health records.
• Next control: Keep adding adversarial synthetic strings to the durable-store contract check before broadening any protected workflow surface.
• Hard stops: PHI introduced, live chart pasted, member ID used, guard weakened without review
• Proof routes: /api/workflows/execution-attempts/durable-store, /health-records, /limitations-workarounds

Strict AAL2 durable-store smoke passed for record, replay, review, and idempotency

• Upgrade: Strict smoke now proves unauthenticated record/replay/review fail closed and authorized AAL2 tenant-admin access can create record, reuse idempotency, replay evidence, and submit review disposition.
• Evidence retained: No-secret pass/fail status, route class, workspace slug, operation classes, and command names; bearer tokens and protected payload details are not retained.
• Verify: npm run smoke:aal2:durable-store:strict
• Rollback/fallback: Turn off SCRIMED_EXECUTION_ATTEMPT_DURABLE_STORE_ENABLED or revoke the tenant role, then verify protected writes fail closed.
• Residual boundary: This is synthetic protected-workflow proof only; it is not production clinical approval, PHI authority, buyer release authority, or SLA evidence.
• Next control: Mirror the strict smoke pattern for every future protected workflow before exposing it in demos, pilots, or buyer diligence.
• Hard stops: buyer release claimed, PHI authority implied, token retained, observer role used for writes
• Proof routes: /workflows/execution-attempts, /release-continuity, /limitations-workarounds

Vercel deploy hygiene hardened around local dependency archives

• Upgrade: The deploy ignore policy, TypeScript exclusion, and ESLint ignore now contain local dependency artifacts including node_modules 2, and production deploys can use archive mode to reduce file-count risk.
• Evidence retained: Deployment identifier, ready status, ignore-file policy, and command class only; no secrets or local cache contents are retained.
• Verify: npm run build
• Rollback/fallback: Remove local generated dependency archives, rerun build, then deploy with the Vercel archive pathway after human release review.
• Residual boundary: A successful lint, build, typecheck, or deploy does not create launch approval, customer release approval, uptime guarantee, or managed-service coverage.
• Next control: Keep generated-output hygiene in predeploy checks and require release notes to distinguish product changes from local artifacts.
• Hard stops: secret in deploy bundle, generated cache deployed, deploy success treated as launch approval, file-count error ignored
• Proof routes: /launch-readiness, /service-reliability, /limitations-workarounds

PHI and live patient-data boundary

• Workaround: Use synthetic fixtures, metadata-only references, external evidence-room pointers, no-PHI excerpts, and Health Records Safety Exchange source mapping.
• Fallback: If a buyer needs live records, convert the ask into a customer-hosted sandbox scoping packet and hold execution until approved authority exists.
• Escalate: Any PHI, identifier, live chart, production credential, patient matching, or data-residency request appears.
• Gate: Executed customer authority, BAA/DPA when required, security review, clinical governance approval, connector approval, monitoring, rollback, and retained evidence.
• Proof routes: /health-records, /clinical-authority-readiness, /boundary-resolution

Live clinical care and CDS authority

• Workaround: Frame outputs as synthetic workflow planning, review queues, draft-only documentation, and clinical-governance preparation with explicit human review.
• Fallback: Route clinical intent, care pathway, diagnosis, treatment, triage, or clinician signature requests to clinical governance before any external claim expands.
• Escalate: A claim implies diagnosis, treatment, live triage, autonomous care coordination, or production CDS.
• Gate: Clinical governance approval, regulatory classification, customer scope, clinician workflow controls, monitoring, override, and rollback evidence.
• Proof routes: /clinical-authority-readiness, /clinical-care-activation, /qa-claim-guard

Production EHR connector and writeback boundary

• Workaround: Use fixture validation, synthetic conformance kits, connector contract review, external artifact references, and no-mutation evidence.
• Fallback: Convert live integration asks into a standards mapping, sandbox preflight, and customer authority checklist.
• Escalate: Production endpoint, credential, writeback, payer submission, patient match, or live HIE/EHR request appears.
• Gate: Customer sandbox, security review, connector contract, data boundary approval, mutation policy, audit, monitoring, and rollback evidence.
• Proof routes: /interoperability, /health-records, /integrations/fixture-validation

Public API, SLA, and unlimited-scale boundary

• Workaround: Use route summaries, boundary headers, version posture, draft quotas, rate-limit classes, support assumptions, and no-SLA language.
• Fallback: If a buyer needs API terms, route to contract review with route class, quota, auth posture, monitoring, incident, support tier, and price floor.
• Escalate: Public API, unlimited usage, uptime, SLA, support, or managed-service language appears.
• Gate: Contract terms, staffing model, support tier, incident response, monitoring, rate limits, price model, and executive approval.
• Proof routes: /platform-power, /enterprise-scalability, /service-reliability

Live AI, production model-routing, and agent autonomy boundary

• Workaround: Use model-route registers, allowed/blocked data classes, eval packs, red-team loops, human approval triggers, and protected AAL2 operator lanes.
• Fallback: When a tool action is requested, switch to recommendation mode and require named human approval before protected execution.
• Escalate: The system is asked to act without approval, route PHI to a model, self-remediate production, or make clinical/legal/financial decisions.
• Gate: Approved provider, model route, data policy, eval pass criteria, monitoring, tool schema, approval UI, rollback, audit persistence, and customer authority.
• Proof routes: /agents, /trust-os, /platform-power, /continuous-review-audit

Security, privacy, SOC 2, HITRUST, and certification boundary

• Workaround: Use Trust Center, Provider Security Reviews, Procurement Evidence Registry, Global Certification Readiness, and no-sensitive-artifact references.
• Fallback: Respond with readiness status, evidence-request owner, due date, and external-review gate instead of certification language.
• Escalate: A buyer asks for SOC 2, HITRUST, ISO, HIPAA certification, pentest, BAA/DPA, vendor-risk approval, or security signoff.
• Gate: Qualified assessment, remediation evidence, approved artifact reference, customer-specific acceptance, and release authority.
• Proof routes: /trust-center, /global-certification-readiness, /pilot-workspace/access

Global legal, privacy, AI Act, GDPR, NHS, MHRA, and regional boundary

• Workaround: Use regional buyer packs, deployment profiles, official-source evidence implications, partner authority registers, and qualified regional counsel review.
• Fallback: Frame global work as no-PHI discovery and localization planning until regional authority exists.
• Escalate: A claim names country launch, public-sector approval, local hosting approval, GDPR/EU AI Act conformity, NHS/MHRA/Australia approval, or government endorsement.
• Gate: Qualified regional legal/privacy/security review, hosting decision, procurement authority, partner approval, and retained release evidence.
• Proof routes: /global-certification-readiness, /global-reach, /deployment-profiles

Legal, finance, accounting, tax, revenue, and profit boundary

• Workaround: Use fixed-scope offers, price floors, buyer-approved measurement plans, counsel review slots, finance/accounting/tax triage, and qualified release authority.
• Fallback: Route exceptions to qualified counsel, finance, accounting, tax, and executive approval before proposal release.
• Escalate: A proposal includes contract approval, non-standard terms, revenue impact, ROI, reimbursement, valuation, fundraising, tax, accounting, or legal conclusions.
• Gate: Qualified review, signed approval, buyer baseline, measurement plan, billing setup, payment terms, and retained release authority.
• Proof routes: /enterprise-business-ops, /growth-engine, /capital-vitality, /public-market-readiness

Email, calendar, demo, meeting, and buyer communication boundary

• Workaround: Use human-reviewed templates, meeting packets, owner handoffs, no-PHI notes, and explicit send/invite approval fields.
• Fallback: Hold the communication in draft state and require sender, recipient, scope, data boundary, and approval evidence before use.
• Escalate: The communication contains PHI, security claims, contract language, pricing exceptions, procurement claims, or clinical promises.
• Gate: Human approval of exact content, recipients, timing, scope, no-PHI boundary, pricing, and follow-up owner.
• Proof routes: /client-onboarding, /offerings, /pilot-deal-room

Protected QA, buyer proof, and release authority boundary

• Workaround: Use AAL2 protected workspaces, no-secret operator packets, QA Completion Bridge, Activation Seal, Proof Promotion, Buyer Proof Release, and release-control runbooks.
• Fallback: When protected proof is requested, run the approved human AAL2 flow and retain only no-secret packet hashes and metadata.
• Escalate: A claim references retained protected proof, customer release, authenticated QA, token handling, evidence-room access, or buyer distribution.
• Gate: Fresh human AAL2 run, no-secret packet, reviewer signoff, release decision, lockbox, recipient authority, access-log reconciliation, and claim guard approval.
• Proof routes: /release-continuity, /qa-manual-execution-console, /buyer-release-control-run, /pilot-workspace/access

UI quality, accessibility, and seamless navigation boundary

• Workaround: Use site navigation, role journeys, Navigation Audit, Product Console, Hub cards, smoke-covered routes, and manual UI review before demos.
• Fallback: If users cannot find a workflow, add it to primary navigation, role journeys, limitation controls, hub views, product actions, and smoke coverage.
• Escalate: A buyer-critical route is hidden, text overlaps, mobile route is unusable, keyboard path is unclear, or accessibility certification is implied.
• Gate: Manual responsive review, keyboard path review, contrast review, copy fit, accessibility remediation evidence, and qualified external review if claims expand.
• Proof routes: /navigation, /product, /hub, /platform-power

Internal innovation, quantum, and future infrastructure boundary

• Workaround: Keep research assigned to the internal research team with private hypotheses, no-public-claim labels, source logs, review owners, and claim-guard blocks.
• Fallback: Translate future research into internal backlog items and remove public wording until validated evidence and qualified review exist.
• Escalate: Quantum, model advantage, cryptographic, clinical, security, financial, or infrastructure superiority language appears in external material.
• Gate: Validated technical evidence, risk review, qualified legal/security review, buyer-safe claim language, and approved release decision.
• Proof routes: /continuous-review-audit, /qa-claim-guard, /boundary-resolution

Daily limitation intake and workaround triage

• Signals: new blocker, new defect, buyer request, claim drift, missing owner, stale route
• Hard stops: owner missing, proof route missing, PHI introduced, unsupported claim repeated

Daily no-PHI and clinical authority review

• Signals: PHI hint, patient identifier, clinical advice claim, record mutation, care pathway language
• Hard stops: live record requested, diagnosis implied, treatment implied, patient matching requested

Daily API, UI, and AI boundary scan

• Signals: new API, UI navigation gap, model route, agent tool request, eval failure, cost spike
• Hard stops: public API SLA implied, live AI implied, agent tool execution without approval, route orphaned

Weekly legal, finance, and deal-risk council

• Signals: pricing exception, margin risk, contract term, ROI claim, tax/accounting question, investor language
• Hard stops: legal advice implied, revenue guaranteed, profit guaranteed, securities material created

Weekly global and certification workaround review

• Signals: certification ask, regional buyer, public-sector path, data residency, hosting profile, procurement evidence
• Hard stops: certified claim, country launch claim, government endorsement, data residency approved early

Pre-demo seamless-navigation review

• Signals: buyer-critical route, primary nav, role journey, hub view, product action, limitation link
• Hard stops: route hidden, text overlap, mobile route unusable, accessibility certification implied

Open limitation pressure

• Evidence: /api/limitations-workarounds
• Lower open pressure is operating discipline, not approval or certification.

Workaround packet coverage

• Evidence: /limitations-workarounds
• Packets are interim controls; they do not become authority without retained evidence.

Hard-stop visibility

• Evidence: /boundary-resolution
• Hard-stop visibility does not waive qualified review.

Proof-route coverage

• Evidence: /navigation
• Proof routes organize evidence; they do not prove protected execution alone.

Known blocker resolution queue

• Evidence: /api/limitations-workarounds
• A resolution queue is operational control, not production, clinical, security, or release approval.

No-secret execution ledger

• Evidence: /api/limitations-workarounds
• Execution ledger entries prove operational control only; they do not retain tokens, PHI, secrets, or buyer-release authority.

Boundary workaround playbook coverage

• Evidence: /api/limitations-workarounds
• A playbook creates operating discipline only; it does not release any preserved boundary.

Boundary preflight fail-closed coverage

• Evidence: /api/limitations-workarounds
• Preflight classification is routing and containment only; it does not process PHI, execute tools, submit payer work, or approve customer go-live.