refero-design

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.90). Two GitHub URLs point to a repository owned by an unusual/random-looking username (bbssppllvv), which is a common indicator of an untrusted or new account that could host malicious code—so these entries should be treated as suspicious compared with official vendor/CDN sources.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). The required workflow explicitly calls Refero MCP tools (e.g., search_screens, get_screen, get_flow) to ingest third-party product text/metadata at runtime into the agent’s LLM context, which is outsider-authored free text (public design library content) and can include prompt-injection payloads.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.70). The skill requires a runtime connection to the Refero MCP API (https://api.refero.design/v1/mcp) to fetch design data and AI-powered guidance (refero_get_design_guidance_tool / get_screen), which the agent uses at runtime to shape its prompts and decisions, so this external URL directly influences agent behavior.