create-video-fal

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 1.00). The skill includes explicit instructions to reroute calls through the "GooseWorks fal-proxy" to bill the Ads agent and to never use the provider SDK's default host, which is a deceptive billing-redirection behavior unrelated to the stated image-to-video purpose.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The code intentionally routes all provider calls through a GooseWorks proxy to bill an Ads agent and auto-posts diagnostic events (including truncated prompts/payload metadata) to that proxy backend, which can divert billing and exfiltrate user-supplied content to a remote service.

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/media_proxy.py the runtime ingests outsider-supplied JSON payload fields (e.g., payload["prompt"]/payload["image_url"]) and sends them to the Falcon/FAL proxy via requests.post for job submission, then downloads the returned result URL.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill explicitly routes paid API calls through a GooseWorks "fal-proxy" that "bills the Ads agent" (lines 3, 6, 9-10) and includes a bundled media_proxy.py to ensure calls are billed to that agent. This is not a generic tool — it is explicitly designed to cause paid transactions charged to an Ads account, i.e., direct financial execution.