TrustDesk is a public, read-only portfolio demonstration. The security posture below is intentional and is part of what the project is meant to demonstrate: mature, secure-by-design judgment rather than a feature-complete production system.

This project uses synthetic data only. No production, customer, deal, employer, or personal data is stored in this repository or the demo environment. Every customer, questionnaire, question, drafted answer, control, and evidence date is invented for demonstration.

• Read-only public demo — no forms that mutate data, no sending, no approvals, no public write endpoints, no admin surface. (Workspace search/filter is purely client-side display state.)
• Synthetic data only — no real customer, deal, employer, or personal data; no PII.
• No client-side secrets — no API keys or sensitive values in browser-delivered code.
• No AI calls in V1 — the confidence score and routing are deterministic (rules and math), which keeps them explainable and auditable.
• Governance gates — answers built on expired/missing evidence are capped, and privacy/AI-governance questions always require human sign-off; nothing auto-sends.
• No file uploads and no uncontrolled public AI prompt box.
• Environment-variable-based secret management — should secrets ever be introduced, they are kept server-side and never prefixed NEXT_PUBLIC_. V1 requires no environment variables.
• TypeScript strict mode and Zod available for input validation.
• Dependency scanning (Dependabot), secret scanning, and static analysis (CodeQL) enabled on the repository.
• Documented threat model — see docs/threat-model.md.

Because this is a synthetic, read-only demo, the security risk surface is minimal. If you do notice a security concern (for example, accidentally committed secrets or a dependency advisory), please open a private security advisory via the repository's Security tab, or open an issue describing the concern without including any sensitive details. There is no SLA; this is a personal portfolio project.