Security Model

Security Model

RunWitness starts with observation, local policy decisions, approval records, sandbox preflight, scoped operator access, and receipts.

Current foundations include:

• opt-in Docker/Podman sandbox execution
• command and network preflight checks
• protected path handling
• rollback evidence
• hashed operator auth
• local secret vault and redaction primitives
• skill trust and permission checks

Current limits:

• normal witnessed commands still execute through host processes unless the operator chooses a container sandbox path
• network egress is not universally blocked outside container runtime modes
• nested agent activity is visible only when the adapter exposes it
• rollback is opt-in and not guaranteed across every failure mode
• hosted multi-user authorization is a foundation, not a complete SaaS trust boundary

Report vulnerabilities privately. Do not post exploit details, secrets, private receipts, or private logs in public discussions.