Policy & Enforcement

Source: app.qpoint.io/security/policy — v2 Figma (Agent Transition)

QControl observes every action through the proxy today. Below is what an enforcing policy would have intervened on — a dry run from data already collected in audit mode.

1 run flagged for dangerous flags →
Enforcement Mode
Violation Handling / Alerting
Details
GlobalMode / Alerting :
Identity & Access
(Slack Message)

Agents must be registered and policy-bound before they run.

(Enforce)
(Slack Message)
Manage

Agents must run under an org-issued identity, not personal accounts.

(Enforce)
(Slack Message)
Manage

Agents may not execute with root or admin privileges.

(Enforce)
Manage
Action & Autonomy
(Audit)

New tools require human approval before agents can call them.

(Audit)
(Text Message)
Manage

Agents may only invoke tools on the approved list.

Beta
(Audit)
(Text Message)
Manage

Agent code execution is confined to a sandbox.

Beta
(Audit)
(Text Message)
Manage
Data Protection
(Slack Message)

Secrets and tokens are redacted before leaving the proxy.

(Enforce)
(Slack Message)
Manage

Agent traffic may only reach approved destinations.

(Enforce)
Manage

Connections below the TLS baseline are refused.

(Enforce)
(Slack Message)
Manage
Input & Output Integrity
(Audit)
(Slack Message)

Inbound prompts are scanned for injection patterns.

Beta
(Audit)
(Slack Message)
Manage

Agent output is checked before it reaches downstream tools.

(Audit)
(Slack Message)
Manage

Models and tools must match pinned, verified sources.

Beta
(Audit)
(Slack Message)
Manage

Qpoint Brand Style Guide