cloudflare-checkbox
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external web pages (Cloudflare challenges) to determine interaction points.
- Ingestion points: Browser accessibility state, screenshots, and DOM content are read by the
mcp__cua_replandcuatools to discover UI controls (SKILL.md). - Boundary markers: The skill explicitly adheres to the computer-use tool's mandatory action-time confirmation policy for CAPTCHAs and requires the agent to ask for user permission before clicking (SKILL.md, Step 3).
- Capability inventory: The skill uses the
mcp__cua_repltool to perform native UI interactions, including clicking the desktop browser's interface. - Sanitization: Instructions mandate inspecting fresh page state to verify the target is a visible, unambiguous checkbox and verifying the outcome after interaction.
- [EXTERNAL_DOWNLOADS]: The skill references external documentation hosted on GitHub.
- Evidence: It links to
https://github.com/jpoindexter/cloudflare-checkbox-skill/blob/main/docs/verification.mdfor verification notes (SKILL.md). This is a vendor-owned resource belonging to the skill author.
Audit Metadata