Verified Hermes build kit · Hospital and clinic administration
Download the kit. Let your own Hermes inspect it. Approve only after the exact activation card.
STEWARD proposes a governed local Mission Control for future AI-assisted administrative preparation across hospitals, health systems, clinics, ambulatory organizations, practices, service lines, and bounded programs. The public package preserves human authority while making mandate, data, uncertainty, organizational coordination, and agent boundaries visible.
Downloading, opening, or unzipping does nothing automatically. A user gives the ZIP to their own Hermes, Hermes performs read-only preflight, shows an exact implementation activation card, and only then may the user approve a local build. Readiness remains Not operational until that user-specific build and runtime evidence complete.
Our Intent
Nurse AI OS is not a replacement for anything you already belong to. Not your unit, not your school, not your professional organization, not your license.
We are not here to compete with existing nursing programs — we’re here to strengthen them.
We are not here to make you rich. Wealth is not the point; integrity is. If value follows good work, good. It follows — it does not lead.
And we are not here to pull you away from the bedside. The opposite. Your clinical practice is the source of your authority. Without it, none of this means anything.
What we are here to do: steward the arrival of AI into nursing practice, so that it happens with nurses at the table rather than around us — and to show, by how we conduct ourselves, what ethical and responsible influence actually looks like.
Stewardship, not disruption. Influence, not extraction.
Current maturity
NAIO will not represent a requirement as enforcement or a declared test as passed evidence. This public artifact is a Hermes functional build kit, not a claim that STEWARD is already installed, institutionally approved, clinically validated, or operational in any user environment.
Preview-state continuity: This is a specification, not software until a user-approved Hermes build completes. The Complete AI OS claim paused language remains the public nonclaim for any environment where runtime not implemented or not yet tested.
Hospital administrator & clinic manager story
This short introduces the future lane for hospital administrators and clinic managers. It is a vision narrative, not evidence that the STEWARD runtime exists, that controls are enforced, or that any organization has authorized operational use.
Self-install artifact
A portable Hermes functional build kit for STEWARD — Hospital & Clinic Administrator Mission Control target version 2.0.0. Its internal manifest declares 12 administrator workspace starters, 24 SuperPowers, 24 workflows, 30 templates, 10 disabled agents, 149 target controls, 200 acceptance checks, and readiness not_operational_build_required.
It is not an installed application, production runtime, institutional deployment, clinical validation, compliance certification, authority grant, connector, schedule, background automation, or permission to process PHI, employee records, incidents, contracts, secrets, or nonpublic financial data.
Why Nurse AI OS matters: it gives administrators a governed rehearsal space for strategy, pilot charters, policy drafts, implementation roadmaps, meeting intelligence, and synthetic operational artifacts before proposals touch live people or systems. An approved SOUL tells Hermes the administrator's mission, governance obligations, communication preferences, risk limits, memory choices, and authority boundaries; STEWARD Mission Control turns them into visible priorities, projects, approvals, evidence, and follow-through.
The conditional companion is for a hospital or clinic administration context only—not generic administration or an automatic role match. If later activated through every separate gate, it could let Hermes work with specifically approved files, forms, documents, and calendars to prepare pilot artifacts. MCP connectors remain disabled with zero exposed tools. Activation remains blocked pending a complete transitive lock and lock-consuming runtime. It cannot process live intake, patient scheduling, clinical operations, workforce data, or institutional records; profile installation and every connector require a separate activation card and fresh approval.
Verify profile SHA-256 before extraction →
Download inactive Pilot-Ops profile →
GIVE-THIS-PACKAGE-TO-HERMES.md, verify the manifest and checksums, and follow the program exactly. Begin with read-only preflight and show me the implementation activation card before making changes. Downloading or unzipping alone does not install anything.”Build-kit SHA-256: c7efccc78b6947466f0e5c48cbd0f1321f9eeecc6ca0b0ce675c07dd4837c306
Architecture kept
The future architecture reserves one proposed population, hospital_clinic_administration, and one proposed namespace, hcadmin_steward.*. These identifiers create no runtime or permission today.
Hospital, health-system, clinic, ambulatory, practice, service-line, program, and bounded-project leaders share recurring needs around mandate, operations, resources, workforce dignity, quality, resilience, and agent governance.
Enterprise executive, hospital administrator, clinic manager, multi-site operator, service-line leader, project lead, dual-hat clinician-administrator, analyst, consultant, trainee, and unknown-authority contexts are never treated as interchangeable.
Title, employment, access, clinical license, technical capability, or prior approval does not establish delegation, threshold, committee role, financial authority, or institutional mandate.
Doctrine kept
Authority before preparation. Context and consequence; Human authority; Authorization and access; Regulation, risk, resources, and records; Trace, transition, expiry, and termination.
A future pass would permit preparation only. It would never confer the underlying authority.
Human-owned organizational coordination. Aim, accountable owner, interdependencies, official systems, governance gates, named receiver, reconciliation, and expiry.
It must not become a shadow command, task, minutes, incident, or policy system.
Bounded agent governance. Objective, human owner, role, risk, data, typed tools and destinations, permission budget, tests, handoff, kill, rollback, purge, retirement, and evidence.
This preview ships zero agents.
Dignity and data
STEWARD must not rank, risk-score, profile, monitor, or predict employees, applicants, clinicians, teams, sites, patients, communities, or vendors. It must not infer sentiment, burnout, impairment, misconduct, protected characteristics, competence, credential status, productivity, performance, staffing adequacy, cause, blame, or reportability.
It must not perform or stage clinical, employment, staffing, credentialing, financial, contracting, legal, policy, security, emergency, incident-closure, or official-record actions. Human approval does not convert a categorically prohibited function into an allowed one.
Florence-X + EDENA build path
Inspect before you trust
The canonical doctrine, one-lane/context-variant architecture, nonclaims, control frameworks, evidence standard, and implementation gates.
Sixteen open enforcement gaps with priorities, risks, required evidence, staged implementation, and the smallest credible next step.
Static archive, checksum, inventory, parity, Office-container audit facts, and the verified build-kit digest. Runtime enforcement is still not established until a user-approved local build executes tests.
Source provenance →
Preview file checksums →
Public manifest →
Download is not installation. Agents propose. Humans judge. Nurses steward.