arcgis-ai-components

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

• Third-party content exposure detected (high risk: 0.80). The skill's documentation (SKILL.md and AGENTS.md) requires the assistant to read web map layer metadata and embeddings from web map item resources and even shows a custom-agent example that fetches external APIs (e.g., https://api.weather.gov/points/...), so agents can ingest third-party web content that may influence tool calls and actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's minimal/runtime example loads and executes remote JavaScript from the ArcGIS CDN (https://js.arcgis.com/5.0/), which auto-registers the AI components and thus directly supplies code that controls agent behavior at runtime.