支持图片生成（SEEDREAM5_0/NANO_BANANA_2）和视频生成（SEEDANCE_1_5_PRO/SORA2），SORA2 支持文生视频和首尾帧图生视频

Before installing or using this skill, consider the following: 1) Required API Key: The SKILL.md and script require AI_ARTIST_TOKEN (and optionally FEISHU_WEBHOOK_URL) but the registry metadata did not declare any required env vars — confirm with the author and do not paste production secrets until satisfied. Create a scoped/test key if possible. 2) Staging endpoint: The code talks to https://staging.kocgo.vip/stage-api/ai — this is a staging/nonstandard endpoint. Understand and trust that service before sending your API key or sensitive prompts. 3) Missing helper files: The docs instruct running scripts/test_config.py and using .env.example, but those files are not in the manifest. Ask the author for the missing files or inspect the repository before running anything. 4) Feishu webhook: If you set FEISHU_WEBHOOK_URL the script will POST generation results to that URL. Only set webhooks you control and review the message format. Treat this as potential data exfiltration vector for generated prompts/results. 5) Review code locally: Because this skill performs outbound network requests and can save files locally, review scripts/generate_image.py (already included) in a controlled environment. Run it with a limited API key and with network monitoring, or in an isolated environment/container. 6) Operational precautions: Do not reuse high-privilege API keys. Prefer a limited/test key, monitor API usage, and confirm the service's privacy/retention policy for prompts and generated media. What would change my assessment: If the registry metadata were updated to declare AI_ARTIST_TOKEN as required (fixing the manifest mismatch), the referenced test_config.py and .env.example were included, and the author explained why a staging host is used (or switched to an official production endpoint), I would raise confidence and could consider the package coherent (likely benign).