ClawHub
Back to skill

Security audit

Power Automate Monitoring

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly coherent for Power Automate monitoring, but it exposes sensitive tenant data and persistent governance changes without enough handling guidance.

Install only if you are comfortable granting tenant-wide FlowStudio cached-store access through FLOWSTUDIO_MCP_TOKEN. Treat trigger URLs as secrets, avoid pasting them into chats or logs, and require explicit user intent before using update_store_flow to change monitoring, notification, criticality, ownership, tier, or security metadata.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (5)

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest repeatedly frames the skill as tenant-wide monitoring, failure analytics, inventory, and aggregated health views, and explicitly contrasts it with runtime-control skills. However, the documented toolset and workflows include write operations via update_store_flow that change flow configuration and governance state. That is broader than a read-focused monitoring/analytics skill as described.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill explicitly documents access to triggerUrl, which for HTTP-triggered flows can function as a bearer-style secret endpoint. Exposing or echoing these URLs without strong handling guidance can enable unauthorized invocation of flows, leakage through logs or chat transcripts, and downstream abuse of connected systems.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
75% confidence
Finding

The early instructions tell the agent to stop using store tools when the first call indicates lack of Pro+ subscription. But the same document later recommends loading the full governance/store-related tool family via tool_search as a general discovery step. This creates conflicting operational guidance about when those tools should be pursued.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

update_store_flow performs writes to governance metadata, notification rules, descriptions, tags, and ownership/support fields, but the markdown does not include a caution that this changes tenant data. Because the skill otherwise focuses heavily on monitoring and read operations, users may not realize this operation is state-changing.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The manifest says this skill should be loaded only for tenant-wide aggregated views and not for debugging a specific run. Yet the documented workflow 'Check a specific flow's health' centers on querying one flow's summary and run history. Although this is not full live debugging, it still expands beyond the stated tenant-wide-only positioning.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.