Baoyu Cover Image

This skill appears to be a coherent cover-image generator, but there are two main concerns to weigh before installing or running it: 1) Safety / copyright guidance in SKILL.md: the instructions explicitly say 'DO NOT refuse to generate' for sensitive or copyrighted figures and require the agent to preserve likenesses from reference images. That encourages the agent to produce stylized likenesses or bypass refusals; if you care about copyright, privacy, or platform content policies, this is a red flag. 2) Data destination ambiguity: the skill says reference images are 'passed to model (usage: direct)' but does not state whether generation runs locally or is sent to a remote API (and if remote, which endpoint). No credentials are declared, so it's unclear how/where image data would be transmitted. Actions you can take: - Ask the publisher or your platform: where is image generation performed (local model, platform API, third-party service)? Are reference images uploaded, and to which endpoint? Is there a content / DMCA / safety policy enforced? - If you proceed, review the EXTEND.md that will be written and choose storage scope 'Project' if you want preferences kept local to a repo. Disable quick_mode to require confirmation each time. - If you must avoid sending user images or likenesses externally, do not use this skill until the backend is explicit and satisfactory. What would change this assessment: explicit declaration of the generation backend (local vs named remote service), clear statement that content policy / refusal behavior is enforced by the platform (not overridden by the skill), and assurance that reference images are handled securely (not exfiltrated) would raise confidence and could move the verdict toward benign.