ClawHub
Back to skill

Security audit

Pipedrive

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed Pipedrive CRM integration that uses Maton-managed OAuth and does not show hidden code, persistence, or unrelated behavior.

Install only if you trust Maton to broker Pipedrive access. Protect MATON_API_KEY, specify the intended Maton connection when multiple Pipedrive accounts exist, and require clear confirmation before creating, updating, deleting, or changing OAuth connections.

SkillSpector

By NVIDIA
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 (4)

Description-Behavior Mismatch

Medium
Confidence
92% confidence
Finding
The skill metadata and description understate the functional scope by claiming focus on deals, persons, organizations, activities, and pipelines, while the document also exposes stages, notes, users, and OAuth connection-management endpoints. This can mislead users or orchestrators into granting or invoking broader capabilities than expected, weakening informed consent and policy enforcement.

Intent-Code Divergence

Medium
Confidence
93% confidence
Finding
The security guidance states access is scoped only to a subset of Pipedrive objects, but the documented API includes additional resource types and connection-management operations. This mismatch can create a false sense of safety and lead agents or users to underestimate what the skill can read, modify, or rebind.

Missing User Warnings

Low
Confidence
79% confidence
Finding
The JavaScript example performs a write operation immediately after a read example without repeating the nearby approval requirement. In agent or copy-paste workflows, this increases the chance that destructive or modifying actions are executed without an explicit user confirmation step.

Missing User Warnings

Low
Confidence
80% confidence
Finding
The Python example demonstrates creating CRM data without repeating a local approval warning, despite earlier guidance that all write operations require explicit approval. This can normalize unsafe automation patterns and cause unintended CRM modifications in agent-driven contexts.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.