AGENT LAB: SKILLS

pipedrive-automation

Warn

Audited by Socket on Feb 16, 2026

1 alert found:

Anomaly
AnomalyLOW
SKILL.md

[Skill Scanner] Skill instructions include directives to hide actions from user BENIGN: The code fragment is a coherent, standard integration guide for Pipedrive automation via a trusted MCP-based toolchain. It describes prerequisites, sequencing, and data flow consistent with legitimate third-party CRM automation workflows. No suspicious or malicious behavior is evident, and credential handling is described in the context of standard OAuth with explicit user authorization flows. LLM verification: Functionally, the skill matches legitimate Pipedrive automation use-cases. However, its architecture centralizes OAuth and all API traffic through a third-party MCP (https://rube.app/mcp) without documenting token handling, retention, scope restriction, or logging practices. This creates a meaningful supply-chain and privacy risk: the MCP operator can access tokens and CRM data, enabling credential theft or data exfiltration if the MCP is untrusted or compromised. The static scan finding (direct

Confidence: 65%Severity: 50%
Audit Metadata
Analyzed At
Feb 16, 2026, 08:36 AM
Package URL
pkg:socket/skills-sh/composiohq%2Fawesome-claude-skills%2Fpipedrive-automation%2F@569c5ea15555d51dc17199b7b87dacfb8c5833ed