SCRIMED Intelligence Control Plane
Executive mission control for governed healthcare intelligence.
One synthetic, verification-first view across work sessions, agents, skills, workflows, context, model policy, approvals, benchmarks, artifacts, capital intelligence, compute resilience, learning, outcomes, cross-platform release evidence, and audit history.
Governance Before Autonomy
Definition of Done, evidence, verification, and human approval determine completion.
SCRIMED Intelligence Control Plane unifies governed agents, skills, workflows, context, routing, verification, benchmarking, artifacts, capital intelligence, voice simulation, compute resilience, learning proposals, outcomes, and audit metadata. It accepts synthetic or de-identified fixtures only and does not authorize autonomous clinical care, PHI processing, EHR writeback, payer submission, investor outreach, securities representations, production deployment, certification claims, or customer activation.
- Runtime authority: SCRIMED Work durable AAL2 control plane
- Mutation authority: AAL2 + tenant RBAC/RLS + durable store + idempotency + approval
- Agent narrative accepted as proof: no
- Hidden chain-of-thought stored: no
Workspace + Active Sessions
Persistent work remains scoped, cancellable, auditable, and rollback-aware.
awaiting_approval
Synthetic Care Coordination Work Session
Prepare a verified synthetic care coordination brief with citations and human review.
Contract
- Maximum steps: 8
- Maximum tools: 10
- Human approval: required
Recovery
- Cancellable: yes
- Rollback: available
- Risk: high
awaiting_approval
Executive Board Brief Work Session
Prepare claims-safe board brief from verified SCRIMED Work telemetry.
Contract
- Maximum steps: 8
- Maximum tools: 10
- Human approval: required
Recovery
- Cancellable: yes
- Rollback: available
- Risk: moderate
Agent OS + Skills Registry
Least-privilege identities expose declared scope, tools, schemas, evaluations, and owners.
enabled-prepare-only
coordinator
Plans bounded work and coordinates specialist handoffs.
Scope
- clinical, operations, research, executive
- Risk ceiling: moderate
- Model class: balanced
Governance
- Owner: Platform Engineering
- Human review: required
- Audit: scrimed-intel-c9d92511
enabled-read-only
clinical-context
Retrieves cited synthetic clinical context without creating clinical authority.
Scope
- clinical
- Risk ceiling: high
- Model class: specialist
Governance
- Owner: Clinical Governance
- Human review: required
- Audit: scrimed-intel-813b575d
enabled-prepare-only
interoperability
Prepares FHIR, HL7, and DICOM metadata mappings and validations.
Scope
- clinical, engineering
- Risk ceiling: high
- Model class: specialist
Governance
- Owner: Interoperability
- Human review: required
- Audit: scrimed-intel-39292fe6
enabled-prepare-only
research
Builds source-grounded research briefs with contradiction and freshness checks.
Scope
- research, clinical
- Risk ceiling: moderate
- Model class: frontier
Governance
- Owner: Research Operations
- Human review: required
- Audit: scrimed-intel-7e66e8ff
enabled-read-only
evidence-validator
Validates schema, citations, source consistency, and evidence freshness.
Scope
- trust-governance, clinical, research
- Risk ceiling: low
- Model class: deterministic
Governance
- Owner: Trust Engineering
- Human review: policy-dependent
- Audit: scrimed-intel-ec8a6a8e
enabled-read-only
clinical-safety
Applies clinical red flags and routes high-risk output to qualified human review.
Scope
- clinical, trust-governance
- Risk ceiling: high
- Model class: specialist
Governance
- Owner: Clinical Safety
- Human review: required
- Audit: scrimed-intel-70eb83bc
enabled-prepare-only
revenue-cycle
Prepares prior-authorization and denial-analysis drafts without submission.
Scope
- revenue-cycle
- Risk ceiling: high
- Model class: balanced
Governance
- Owner: RCM Governance
- Human review: required
- Audit: scrimed-intel-73f7766a
enabled-prepare-only
patient-access
Prepares referral and scheduling workflow recommendations without outreach or booking.
Scope
- patient-access, operations
- Risk ceiling: high
- Model class: balanced
Governance
- Owner: Patient Access
- Human review: required
- Audit: scrimed-intel-8c91b451
enabled-prepare-only
documentation
Prepares structured documentation drafts for review.
Scope
- clinical, operations
- Risk ceiling: high
- Model class: balanced
Governance
- Owner: Clinical Documentation
- Human review: required
- Audit: scrimed-intel-3be85cae
enabled-prepare-only
artifact-writer
Creates cited Markdown and JSON artifacts from verified evidence.
Scope
- clinical, research, executive, revenue-cycle
- Risk ceiling: moderate
- Model class: balanced
Governance
- Owner: Product Operations
- Human review: required
- Audit: scrimed-intel-e737832f
enabled-prepare-only
executive-intelligence
Prepares claims-safe executive and board briefs.
Scope
- executive
- Risk ceiling: moderate
- Model class: frontier
Governance
- Owner: Executive Operations
- Human review: required
- Audit: scrimed-intel-be183576
enabled-prepare-only
capital-intelligence
Scores investor fit and prepares internal diligence drafts without outbound action.
Scope
- capital-intelligence, executive
- Risk ceiling: high
- Model class: frontier
Governance
- Owner: CEO + Capital Strategy
- Human review: required
- Audit: scrimed-intel-0ea17cff
enabled-read-only
infrastructure-observer
Assesses provider capacity, residency, concentration, and fallback readiness.
Scope
- engineering, trust-governance
- Risk ceiling: moderate
- Model class: deterministic
Governance
- Owner: Platform Reliability
- Human review: required
- Audit: scrimed-intel-ec0093c4
enabled-read-only
policy-reviewer
Applies safety, privacy, authorization, and claims policies before completion.
Scope
- trust-governance, clinical, executive, capital-intelligence
- Risk ceiling: high
- Model class: deterministic
Governance
- Owner: Trust + Compliance
- Human review: required
- Audit: scrimed-intel-d58add6c
Workflow Engine
Reusable task graphs stop at verification and human gates before consequential action.
patient-access
Patient Intake Preparation
Prepare a verified patient intake preparation artifact for human review.
Agents
coordinator, patient-access, documentation, evidence-validator, policy-reviewer
Boundary
External actions enabled: no
patient-access
Referral Coordination
Prepare a verified referral coordination artifact for human review.
Agents
coordinator, patient-access, clinical-context, evidence-validator, policy-reviewer
Boundary
External actions enabled: no
revenue-cycle
Prior Authorization Preparation
Prepare a verified prior authorization preparation artifact for human review.
Agents
coordinator, revenue-cycle, clinical-context, evidence-validator, policy-reviewer
Boundary
External actions enabled: no
revenue-cycle
Denial Analysis
Prepare a verified denial analysis artifact for human review.
Agents
coordinator, revenue-cycle, evidence-validator, policy-reviewer
Boundary
External actions enabled: no
clinical
Clinical Documentation Preparation
Prepare a verified clinical documentation preparation artifact for human review.
Agents
coordinator, documentation, clinical-safety, evidence-validator, policy-reviewer
Boundary
External actions enabled: no
clinical
Patient Education Drafting
Prepare a verified patient education drafting artifact for human review.
Agents
coordinator, clinical-context, artifact-writer, evidence-validator, policy-reviewer
Boundary
External actions enabled: no
research
Research Brief Generation
Prepare a verified research brief generation artifact for human review.
Agents
coordinator, research, artifact-writer, evidence-validator, policy-reviewer
Boundary
External actions enabled: no
executive
Executive Reporting
Prepare a verified executive reporting artifact for human review.
Agents
coordinator, executive-intelligence, artifact-writer, evidence-validator, policy-reviewer
Boundary
External actions enabled: no
capital-intelligence
Investor Diligence Preparation
Prepare a verified investor diligence preparation artifact for human review.
Agents
coordinator, capital-intelligence, executive-intelligence, evidence-validator, policy-reviewer
Boundary
External actions enabled: no
Context Fabric + Multi-Model Router
Cited context is ranked as untrusted data; policy selects a model class without making a provider call.
reviewed-reference
Care coordination review SOP
Care coordination briefs require source citations, unresolved risk flags, and human review before patient-facing use.
Ranking
- Unified: 66
- Ontology: 54
- Graph: 50
- Freshness: 90
Citation
SCRIMED synthetic SOP CC-001
synthetic-fixture
Prior authorization draft documentation rule
Prior authorization packets must identify missing symptom language, functional status, visit timing, and medical necessity support before submission.
Ranking
- Unified: 45
- Ontology: 18
- Graph: 65
- Freshness: 90
Citation
SCRIMED synthetic payer-rule PA-004
reviewed-reference
Board brief evidence template
Board briefs must separate verified evidence, assumptions, unresolved risks, and next approvals.
Ranking
- Unified: 42
- Ontology: 36
- Graph: 50
- Freshness: 90
Citation
SCRIMED synthetic executive template BB-002
reviewed-reference
FHIR R4 preview contract
FHIR bundle previews are read-only and never write to an EHR without future connector approval.
Ranking
- Unified: 41
- Ontology: 18
- Graph: 65
- Freshness: 90
Citation
SCRIMED synthetic FHIR preview FP-001
SOL_CLASS
Model routing explanation
SOL_CLASS selected from complexity, consequence, verification strength, privacy, residency, latency, and budget. Selected the strongest policy-compliant provider that preserves privacy, residency, budget, and task capability requirements.
Selection
- Provider class: frontier
- Model: scrimed-synthetic-no-call
- Human review: required
Execution
Provider calls executed: no
Verification + Reasoning Observatory
Operational evidence is visible without exposing hidden chain-of-thought.
review-required
Verification result
pause_for_review
Trust
- Total: 85
- Evidence coverage: 100
- Policy compliance: 100
Failed checks
human-approval-state
within-scope
Objective drift monitor
continue-within-definition-of-done
Current step
Verify citations and unresolved limitations.
Drift flags
none
ConsequenceBench + Artifacts
High-consequence, worst-group, edge-case, lineage, and staleness evidence outrank generic leaderboard scores.
awaiting-evaluation-results
Consequence-weighted benchmark
Synthetic evaluation infrastructure only; zero scores mean no evaluation evidence was supplied, not model failure or clinical validation.
Coverage
- Cases: 21
- High-consequence performance: 0
- Worst-group performance: 0
Distribution shift
requires-approved-shift-dataset
current-metadata
Artifact lineage
retain-current-review-status
Artifacts
1
External use
human approval required
Capital Intelligence + Compute Resilience
Internal opportunity analysis and infrastructure resilience stay evidence-led and approval-gated.
internal-research-and-drafting-only
Capital Intelligence Copilot
Internal research, qualification, analysis, and draft preparation only. All outbound activity requires CEO approval.
Profiles
3
Outbound
disabled; CEO approval required
review-required
cloud
Compute resilience score: 70
Fallback
Policy reevaluation: required
Migration
Silent protected migration: blocked
review-required
customer-vpc
Compute resilience score: 71
Fallback
Policy reevaluation: required
Migration
Silent protected migration: blocked
blocked-for-hardening
air-gapped
Compute resilience score: 59
Fallback
Policy reevaluation: required
Migration
Silent protected migration: blocked
review-required
edge-device
Compute resilience score: 66
Fallback
Policy reevaluation: required
Migration
Silent protected migration: blocked
Approval Achievement
Evidence, dependencies, owners, expiry, and authority stay explicit before scope expands.
one-automated-technical-gate-achieved-human-and-external-approvals-pending
Approval critical path
An automated technical gate is not legal, privacy, clinical, regulatory, security-assurance, buyer, PHI, production, or customer go-live approval.
Completed scope
- Automated technical gates: 1
- Human/external approvals: 0
Next candidate
SCRIMED Intended Use Internal Approval
achieved-automated-technical
No-PHI Control Plane Automated Technical Quality Gate
Provides reproducible technical evidence for demos and diligence without implying production, clinical, privacy, or regulatory approval.
Missing evidence
none within automated technical scope
Next action
Rerun the full gate after material source changes and before every protected pilot release candidate.
evidence-complete-awaiting-human-signoff
SCRIMED Intended Use Internal Approval
Reduces claim inconsistency, shortens buyer review, and protects credibility across every sales and investor channel.
Missing evidence
Founder/CEO approval, qualified legal review, clinical governance review, versioned effective date and review expiry
Next action
Review and sign the proposed memo, record reviewer identities and dates outside source code, then align public copy and sales assets to the approved version.
evidence-build-active
No-PHI Protected Pilot Internal Release
Creates the fastest defensible path to paid pilots, referenceable outcomes, and enterprise procurement evidence.
Missing evidence
approved Intended Use Memo, buyer scope and success metrics, named reviewer signoffs, strict AAL2 protected-write smoke, no-PHI canary evidence
Next action
Complete Intended Use approval, strict nonproduction AAL2 smoke, buyer scope, reviewer assignments, and no-PHI canary packet.
external-review-required
HIPAA/BAA and Production PHI Readiness
Required for many health-system enterprise workloads, but not needed to sell synthetic assessments and no-PHI pilots now.
Missing evidence
risk analysis, policy and safeguard evidence, BAA/DPA path, vendor review, customer authorization
Next action
Commission a formal HIPAA readiness assessment while continuing to prohibit PHI in current product flows.
external-review-required
SOC 2 Independent Assurance
High buyer-confidence leverage for Microsoft-, NVIDIA-, OpenAI-, Apple-, Micron-, payer-, and health-system-grade diligence.
Missing evidence
readiness assessment, control remediation, evidence window, external examination
Next action
Select a readiness assessor, scope Security/Availability/Confidentiality first, and begin evidence collection without claiming attestation.
external-review-required
FDA CDS/Device Classification
Protects differentiation and clinical credibility while preventing accidental regulated-device claims.
Missing evidence
qualified classification memo, module claims map, clinical evidence plan if regulated
Next action
Keep current functions decision-support/draft-only and obtain a module-level classification memo before clinical claims expand.
external-review-required
ONC/HTI and Connector Certification Decision
Turns FHIR, HL7, DICOM, X12, RIS, HIS, and PACS capability into credible integration evidence without overclaiming certification.
Missing evidence
one selected connector target, certification applicability memo, sandbox acceptance, production monitoring and rollback
Next action
Select one revenue-linked connector, define whether certification applies, and complete a synthetic customer sandbox acceptance plan.
external-review-required
EU AI Act/GDPR Regional Release
Creates a defensible path to European health-system and life-sciences buyers without premature conformity claims.
Missing evidence
current legal classification, DPIA, data transfer mechanism, regional deployment profile, buyer approval
Next action
Run a no-personal-data EU discovery assessment and commission current legal classification before regional launch language.
external-review-required
UK NHS DTAC/MHRA Buyer Release
Supports NHS and private-provider diligence across clinical safety, security, privacy, interoperability, accessibility, and usability.
Missing evidence
DTAC evidence, clinical safety file, privacy/security review, buyer-specific acceptance
Next action
Build a product-specific DTAC evidence index and obtain UK clinical-safety classification before buyer submission.
external-review-required
Australia Essential Eight and Health Procurement Release
Builds a credible regional path through cyber maturity, hosting, privacy, procurement, interoperability, and buyer governance.
Missing evidence
maturity mapping, privacy/hosting review, regional legal review, buyer-specific acceptance
Next action
Choose a target buyer profile and map current controls to the required Essential Eight maturity and health procurement evidence.
evidence-build-active
FaithCore Faith-Health Pilot Governance
Differentiates SCRIMED through respectful institution-specific governance while retaining clinical, privacy, accessibility, and non-discrimination standards.
Missing evidence
faith organization charter, clinical governance signoff, patient-choice safeguards, buyer approval
Next action
Prepare one organization-specific FaithCore governance charter and synthetic workflow workshop for named human review.
evidence-build-active
Buyer-Specific Proof Release
Converts platform proof into sales velocity while preserving confidentiality, claims integrity, and recipient control.
Missing evidence
named recipient, approved packet scope, reviewer signoffs, access control, post-share reconciliation
Next action
Complete one buyer-specific release packet through the protected runbook without public distribution.
Cross-Platform Evidence
GitHub, Vercel, Supabase, Wix, and Figma evidence is reconciled before release claims expand.
release-evidence-drift-blocks-promotion
Release evidence decision
This snapshot reconciles no-secret operational metadata only. It grants no deployment, PHI, clinical, payer, EHR, customer, testimonial, regulatory, certification, or go-live authority.
Current blockers
4
Production promotion
blocked
Proof Before Promises
Inspect the evidence behind SCRIMED.
Review no-PHI demonstrations, governance controls, interoperability evidence, and clearly bounded pilot-readiness artifacts before making a buying decision.
Primary action
Review the Trust Center
Public disclosure
Synthetic demonstration and decision-support scope only. Do not submit patient information through this website.
blocked · refresh-required
github · source-control
The public main branch is behind the current local SCRIMED build, and the working tree contains a large unreviewed change set.
Remediation
implemented-local: scripts/release-provenance-preflight.mjs, .github/workflows/ci.yml, docs/release-provenance.md
Next action
Partition the related working-tree changes, run the complete quality gate, and prepare one reviewable release candidate without committing unrelated state.
blocked · refresh-required
vercel · deployment
The latest production deployment is READY, but its metadata identifies a dirty source tree and the current control-plane endpoint is not present on the custom domain.
Remediation
implemented-local: SCRIMED_RELEASE_PROVENANCE_ENFORCED, SCRIMED_APPROVED_RELEASE_SHA, npm run release:provenance:strict
Next action
Block further promotion until a clean immutable revision passes CI, deploys through the approved pipeline, and the intended live route set passes public smoke.
blocked · refresh-required
supabase · data-plane
The protected-pilot project is healthy, but local durable-store and evidence-binding migrations are not present in the observed remote migration history.
Remediation
prepared-not-applied: supabase/migrations/20260711224500_stored_vector_advisor_index_hardening.sql, docs/supabase-advisor-remediation.md
Next action
Create or authorize a nonproduction branch, apply pending migrations there, resolve advisor findings through reviewed migrations or documented risk acceptance, and run strict AAL2 smoke.
blocked · refresh-required
wix · public-claims
The published marketing site contains a named client testimonial for which the repository approval graph has no authorization evidence.
Remediation
external-action-required: scripts/public-claims-integrity-smoke.mjs, docs/public-claims-integrity.md, Voice Intake Assistant form disabled and verified at revision 2, Proof Before Promises replacement copy
Next action
Remove the named testimonial, label every Wix collection path no-PHI, and replace social proof with evidence-based platform proof until signed customer authorization and claim substantiation exist.
warning · refresh-required
figma · design-governance
A Figma account is connected with a view seat, but no canonical SCRIMED design file is linked to this release evidence snapshot.
Remediation
implemented-local: docs/scrimed-design-source-of-truth.md, app/globals.css
Next action
Name one canonical design file, grant the appropriate design ownership, and bind page-level visual acceptance evidence to release review.
Voice + Learning + Outcomes + Audit
Simulation, reviewed corrections, baseline-first value measurement, and immutable evidence close the loop.
simulation-only
Voice workflow
Voice workflow is simulation-only and does not diagnose, treat, prescribe, triage emergencies, outreach patients, or store raw audio.
States
idle → listening → processing → speaking → ended
Raw audio stored
no
review-gated
Learning proposals
Memory records validated facts and decisions; learning proposals change behavior only through tests, approval, and normal deployment.
Proposals
1
Self-deployment
blocked
baseline-first-no-fabricated-improvements
Outcome and value intelligence
No outcome improvement is claimed until approved baseline, measurement, attribution, reviewer, and statistical evidence exist.
Metrics
20
Post values present
no
metadata-only
Audit history
Initiator, objective, model, context, tools, policy, approvals, verification, cost, latency, state, and rollback remain traceable without raw PHI, secrets, prompts, or chain-of-thought.
Events
1
Audit hash
scrimed-intel-339049f3