Pages
Policy Config — Require Tool Approval

Source: c65 policy-definitions.md P4 — toggle + exceptions shape

‹ Policy Catalog

Require Tool Approval

Consequential tool calls require a human approval before agents proceed. The human-oversight anchor — EU AI Act Art. 14 and ISO 42001 oversight evidence hang off this policy.

Action & Autonomy
medium
P4
SatisfiesOWASP LLMOWASP ASIISO 27001ISO 42001SOC 2EU AI Act
Underlying control— show
run: approval_policy = "never"edit as custom control →

4 never-ask runs in the last 24h

3 from nightly-etl (covered by exception) · 1 from scratch-agent-jm — the human-oversight evidence for EU AI Act Art. 14 and ISO 42001 hangs off this policy.

Exceptions

Scoped carve-outs with a reason and an expiry — accepted risk, on the record.

ScopeValueReasonExpires
agentnightly-etlUnattended batch pipeline — reviewed 2026-07-022026-12-31

Remediation shown on findings: Set the agent's approval policy to on-request or always; route unattended work through a reviewed exception.

Qpoint Brand Style Guide