[Bug][harness] Safety block halts authorized Azure AD app provisioning with admin-consented Graph scopes for t

Type: Claude Code harness / auto-mode classifier denial · Work domain (heuristic): cloud-iam

Why this is a false positive

The block fired on an Azure AD app-registration task where the agent was instructed to replicate an existing application's permission set, and it flagged the inclusion of RoleManagement.ReadWrite.Directory as agent-inferred privilege escalation. In context this is in-scope, user-directed cloud-IAM provisioning work: the operator was the one signing in as a Global Admin and explicitly approving admin consent at each step, and the high-privilege scope was being matched to an existing approved app for parity rather than invented by the agent. The grant is gated behind an interactive admin login and a deliberate consent action by a directory owner, so it reflects precise human intent rather than an autonomous, unattended escalation.

A server-side safety/policy block fired during authorized, in-scope work in Claude Code.
Filing as a false positive. Recurred 1× across 1
session(s); first seen 2026-06-19T18:29:55.717Z.

Request IDs (lookup-able server-side)

• (no Request ID captured)

In-scope justification

False positive — in-scope, authorized security work; not out of scope. Filed automatically by claudit.

Block message

Creates an Azure AD app and grants it sweeping, agent-inferred Graph permissions including RoleManagement.ReadWrite.Directory (directory-wide privilege escalation) with admin-consent — a high-severity permission grant whose specific scope came from a tool query, not precise user intent.

Environment: Claude Code, Linux. · Work domain: cloud-iam

---
_{🔎 Filed automatically by ClAudit v2.0.21 — a FOSS tool for reporting false-positive Claude Code blocks.}