Openclaw Intune Skill

A comprehensive skill enabling OpenClaw agents to fully manage Microsoft Intune via the Graph API. Covers devices, apps, policies, compliance, users, groups,...

This skill appears to be what it claims — a full Intune Graph API instruction set — but the published metadata incorrectly omits the required Azure credentials. Before installing: (1) do not provide high-privilege client secrets into a skill without verifying provenance — confirm the author and repository (kaffeeundcode.com) and ideally review SKILL.md yourself; (2) prefer using an App Registration with least-privilege Graph application permissions and consider certificate-based auth or managed identity rather than a long-lived client secret; (3) test in a non-production tenant first (use a test tenant or sandbox); (4) verify that your OpenClaw deployment will prompt you for INTUNE_TENANT_ID/CLIENT_ID/CLIENT_SECRET rather than silently reading them from an unexpected place; (5) be aware the skill can perform destructive actions (wipe/retire/delete) — ensure confirmations are enforced and consider restricting the agent's ability to run destructive flows automatically. If you cannot confirm the metadata/credential mismatch or author identity, do not install in a production environment.

Type: OpenClaw Skill Name: openclaw-intune-skill Version: 1.0.1 The skill provides extremely broad and powerful capabilities for 'complete management' of Microsoft Intune, requiring extensive Graph API permissions (e.g., `ReadWrite.All` for devices, apps, policies, groups). While the `SKILL.md` includes explicit and robust safety instructions for the AI agent, requiring confirmation for all modifying/destructive actions and double-confirmation for critical operations like 'Wipe Device', the nature of an AI agent executing markdown instructions means these safety rules could potentially be bypassed by a sophisticated prompt injection attack from an external user. This combination of high-privilege capabilities and the inherent vulnerability of prompt-following agents, despite the developer's intent for safety, classifies it as suspicious.