chrome-extension-dev
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is primarily educational, providing detailed documentation and best practices for modern browser extension development. It explicitly warns against dangerous practices like remote code execution and the use of
eval(), which are prohibited in Manifest V3 for security reasons. - [INDIRECT_PROMPT_INJECTION]: The guide describes a 'ref number system' for browser automation in
references/patterns.md. This architecture allows an AI agent to identify and interact with DOM elements. While this is a standard pattern for agent-based web interaction, it acknowledges an attack surface where untrusted web content (HTML elements/attributes) is processed by the agent. The skill mitigates this by focusing on legitimate automation and adhering to Manifest V3 security constraints. - Ingestion points: The
assignRefNumbersfunction inreferences/patterns.mdreads interactive DOM elements from active web pages. - Boundary markers: None explicitly defined in the architectural snippet; implementation is left to the developer.
- Capability inventory: The described system allows for
clickandfocusactions on web elements. - Sanitization: The skill promotes Manifest V3 compliance, which naturally restricts the execution of untrusted code injected via the DOM.
Audit Metadata