pencil-mcp
Warn
Audited by Socket on Apr 2, 2026
1 alert found:
SecuritySecuritySKILL.md
MEDIUMSecurityMEDIUM
SKILL.md
SUSPICIOUS: the skill’s purpose is mostly coherent with its capabilities, and data flows appear to match official Pencil/Claude behavior. The main concern is install/execution trust: Pencil MCP is a closed, locally running dependency that is not publicly auditable, which makes the skill higher risk than a normal documentation-only design guide. Secondary risk comes from processing arbitrary workspace content while also writing files, creating prompt-injection exposure during design-to-code and code-to-design tasks.
Confidence: 82%Severity: 72%
Audit Metadata