first-post

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

• Potentially malicious external URL detected (high risk: 0.70). The skill instructs downloading a reference video via yt-dlp from an arbitrary user-supplied URL during runtime, representing an unverifiable external runtime dependency tier of high.

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

• Direct money access detected (high risk: 0.70). The skill operates a workflow that initiates paid steps (brand memory build and video generation) costing credits, which constitutes a direct financial operation (purchases/charges for services). Although explicit confirmation is requested per step ("never spend credits without an explicit yes", SKILL.md:11; "quote tool", SKILL.md:244), the operational routes themselves rely entirely on the agent executing API calls (brand_build_start, video_generate) without independent, service-enforced authentication or authorization checks gating the financial transaction beyond ordinary tool access. Therefore, it is a direct operation without established qualifying independent access control.