Skills
Navigate
skills/mapbox/mapbox-agent-skills/mapbox-store-locator-patterns/Gen Agent Trust Hub

mapbox-store-locator-patterns

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill demonstrates patterns that ingest GeoJSON data from potentially untrusted external sources (such as an API endpoint) and interpolates property values into the UI.
  • Ingestion points: SKILL.md (Step 3), AGENTS.md, and references/search-filter.md process properties from GeoJSON features.
  • Boundary markers: No delimiters or safety instructions are used to distinguish untrusted data from UI code.
  • Capability inventory: The code uses Mapbox's setHTML() method and standard DOM innerHTML to display data.
  • Sanitization: The skill lacks sanitization, escaping, or validation logic for location properties (like name, address, or phone) before rendering them as HTML, creating a vulnerability to data-poisoning attacks.
  • [DYNAMIC_EXECUTION]: Implementation examples in SKILL.md, AGENTS.md, and references/geolocation-directions.md use .setHTML() on mapboxgl.Popup instances and .innerHTML on sidebar listings. This allows for the dynamic execution of arbitrary HTML and JavaScript if malicious content is embedded in the GeoJSON data provided to the skill.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 09:08 AM