Security — Built secure, not bolted on

1. Security Architecture 2. Read-Only Scanning 3. Credential Management 4. Data in Transit 5. AI Execution Security 6. Access Control 7. Instance Hygiene 8. Incident Response 9. Responsible Disclosure 10. Customer Recommendations

01 — Architecture

Security Architecture Overview

GeneWorks is designed as a client-proxied, credential-local, read-first platform. The two most sensitive aspects of any ServiceNow automation tool are the handling of instance credentials and the ability to write changes. GeneWorks resolves both by construction: it stores no credentials on its servers, and it makes no change to your instance without an explicit human action.

When you connect an instance, credentials are stored only in your browser and API calls are constructed client-side and sent directly to your instance. At no point do credentials pass through GeneWorks infrastructure, and at no point does a scan write to your instance.

// Trust boundaries — where data lives and moves Your browser credentials (localStorage) · session state │ HTTPS · TLS 1.2/1.3 · direct ▼ Your ServiceNow your records stay here · scans read in place · never copied off ▲ │ read-only scans · writes ONLY after human sign-off GeneWorks agent orchestration · workspace memory (summaries only) ✗ no credentials ✗ no copy of your records

Principle	How it's enforced
No server-side credentials	Credentials live only in your browser; GeneWorks servers never receive them.
No copy of instance data	Scans return counts, scores, and findings — summaries of the shape of your instance, not your records.
Read-only scanning	All eight apps read; none write on their own.
Session-scoped execution	No background jobs, no persistent connection between sessions.
Reviewed, reversible changes	A change leaves the loop only after human sign-off; every write is captured for backout.
Never production	Sub-production only; production is blocked at the connection test.

02 — Read-Only Scanning

Read-Only Scanning & Remediation

Every GeneWorks app follows the same shape — scan → score → propose → you approve. The scan is strictly read-only: it reads sys tables, dictionary, audit, and configuration records to build its map, health score, or findings, and writes nothing back.

// The read/write boundary is explicit per app App Reads Writes Command Center tables, relationships, sys_metadata none Application Explorer app artifacts, customizations none CMDB Assessment cmdb_ci, dictionary, sys_audit none Catalog Center sc_cat_item, variables, client scripts none Normalization core_company (read), reference DB none until "Apply" UpgradeAssist release notes × instance, sys_upgrade… none Delivery Loop target tables (read-before-write) only after sign-off

Remediation is a separate, deliberate step. When you accept a finding — "Make it a Change" in CMDB Assessment, "Propose change" in Catalog Center, "Apply to instance" in Normalization — GeneWorks opens a governed change that runs through the same human-gated loop as any build. Nothing is applied silently, in bulk, or without a named approver.

The assessment tool is also the remediation tool — but never at the same moment. Diagnosis is read-only and instant; remediation is explicit, reviewed, and reversible.

03 — Credentials

Credential Management

Your ServiceNow credentials — instance URL, username, and password (or OAuth token) — are stored in your browser. This storage mechanism is:

Scoped to your origin: browser storage is strictly origin-scoped. No other website or tab can read data stored by GeneWorks.
Never transmitted to GeneWorks servers: credentials are used only to construct direct API calls to your instance.
Cleared on demand: remove credentials at any time via Settings or by clearing browser storage — an immediate, permanent operation.

Integration credentials (OAuth tokens, API keys, Basic Auth used by an integration GeneWorks builds) are created as records in ServiceNow's native encrypted Credential Store. GeneWorks creates the record but never reads the secret back.

Use a dedicated service account. A gw_service_account gives you a clean audit trail and lets you revoke GeneWorks access independently of other accounts. Start it read-only.

04 — Data in Transit

Data in Transit

All data transmitted between GeneWorks and external services is encrypted in transit. GeneWorks does not transmit any data over unencrypted HTTP, and refuses to connect to an instance endpoint that does not serve a valid HTTPS certificate.

Channel	Protocol	Carries
Browser → ServiceNow	HTTPS · TLS 1.2/1.3	All REST API calls, authenticated with your credentials.
Browser → AI model API	HTTPS	Task prompts. Managed tier: the latest Claude Opus. BYO tier: your own model.
Contact form	HTTPS	Form submissions forwarded to our designated address.

If your ServiceNow instance is not accessible via HTTPS, GeneWorks cannot connect to it.

05 — AI Execution Security

AI Execution Security

The GeneWorks agent workforce interacts with your instance only within the boundaries of the task you explicitly submit, under strict architectural constraints.

Read-before-write: before creating or modifying any artifact, the responsible agent reads the current state of the target — the same discipline the scanning apps use everywhere. This reduces the risk of overwriting existing logic.
No autonomous background execution: agents do not execute against your instance outside an active session. No scheduled jobs, no webhooks that trigger agents without you, no persistent connection between sessions.
Prompt-injection resistance: system instructions are separated from user input. Content in an uploaded document, a catalog item, or a scanned record cannot override agent behaviour, security constraints, or governance rules.
Bounded, orchestrated roster: 152 specialist agents work under Gene's orchestration, each scoped to its discipline; the Normalization reference DB is maintained by its own 18 agents. No agent escalates beyond the connected service account's permissions.

// Defences applied to every agent message and action Security fence injected server-side into every agent context before the model sees it; not modifiable at runtime. Blocked patterns mass deletes · production-table truncation · credential modification → refused at the execution layer Scoped writes every write lands in a named Update Set, never Default.

06 — Access Control

Access Control & Permissions

GeneWorks operates within the permission model of your instance. Agents can only perform actions the provided service account is authorised to perform. GeneWorks does not attempt to escalate privileges or circumvent ACLs.

Least privilege, staged by phase. Grant only what the work in front of you needs:

Phase	Recommended role	Why
Assessment (read-only)	read-scoped account	Command Center, CMDB Assessment, UpgradeAssist, Catalog Center, Explorer all scan read-only.
ITSM remediation	itil_admin	Sufficient for most ITSM builds and changes.
Platform work	admin	Needed for ACLs, roles, and platform-wide changes.

For multi-user workspaces, GeneWorks provides workspace-level user management scoped to your workspace session.

07 — Instance Hygiene

ServiceNow Instance Hygiene Recommendations

To maintain the highest security posture when using GeneWorks:

Use a dedicated GeneWorks service account with the minimum required roles — never share credentials with other administrators.
Start read-only. Point the assessment apps at an instance with a read-scoped account first; grant write roles only when you move to remediation.
Enable audit logging on the service account for a full record of every API call GeneWorks makes.
Review every change in a non-production environment before promoting to production — GeneWorks packages the change; human review before production commit remains your responsibility.
Set session timeout policies on the service account appropriate to your organisation's requirements.
Monitor the sys_audit table for GeneWorks service-account activity, and revoke the account immediately if you suspect misuse.

08 — Incident Response

Security Incident Response

In the event of a security incident affecting GeneWorks infrastructure or customer data, we commit to the following timeline:

Stage	Timeframe	What happens
Initial assessment	≤ 24 hours	Scope, severity, and nature of the incident assessed.
Customer notification	≤ 72 hours	Affected customers notified of nature, data potentially affected, and steps taken or planned.
Remediation	Ongoing	Contain, remediate, prevent recurrence — with progress communicated.
Post-incident review	≤ 30 days	Report made available to affected Enterprise customers.

Structurally limited blast radius. Because credentials are stored client-side, no instance data is held on our servers, and scanning is read-only, a compromise of GeneWorks servers cannot directly expose your ServiceNow credentials or data.

09 — Responsible Disclosure

Responsible Disclosure Policy

We welcome security researchers to responsibly disclose vulnerabilities discovered in the GeneWorks platform. If you discover a potential vulnerability, please:

Email a detailed description with the subject line "GeneWorks Security Disclosure".
Include steps to reproduce, potential impact, and any proof-of-concept code.
Allow us a reasonable timeframe (90 days) to investigate and remediate before public disclosure.
Do not access, modify, or exfiltrate customer data during your research.
Do not perform denial-of-service attacks, social engineering, or physical security tests.

We acknowledge receipt within 5 business days and will not pursue legal action against researchers who follow this policy in good faith. Found a security issue? Reach out directly at gene@geneworks.ai — you'll hear back within 5 business days.

10 — Customer Recommendations

Security Recommendations for Customers

Browser security: use GeneWorks only on trusted, organisation-managed devices with up-to-date browsers. Avoid shared or public computers.
Session management: log out when your session is complete, especially on shared workstations.
Password strength: use strong, unique credentials for your GeneWorks service account. Do not reuse credentials across systems.
Production caution: always test GeneWorks-generated changes in a non-production instance before committing to production. GeneWorks provides ATF and functional coverage, but it cannot test scenarios not described in your requirements.
Access reviews: periodically review the GeneWorks user list in your workspace and remove access for team members who no longer require it.
Stay updated: monitor the GeneWorks website for security advisories. Material updates are communicated directly to registered customers.

Security is the architecture, not a policy promise. Read-only scanning, credential-local storage, sub-production-only connections, and two human gates are enforced by construction — not left to good intentions.