laniameda-gallery-query

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

• Third-party content exposure detected (high risk: 0.90). This skill queries the laniameda.gallery Convex backend and can list public assets (scope='public'), return asset.promptText and sourceUrl, and download external URLs (see SKILL.md and scripts/query.ts), and the docs explicitly instruct agents to "use the savedPath and promptText in the current task," meaning untrusted/user-ingested third-party prompt content can be read and influence subsequent actions.

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

• Potentially malicious external URL detected (high risk: 0.80). The script requires a runtime CONVEX_URL and makes POST requests to ${CONVEX_URL}/api/query and ${CONVEX_URL}/api/action to fetch asset data (including promptText) that is injected into the agent workflow, so remote content can directly control prompts.