Live2D / VRM / MMD 形象渲染这一层的不变式和踩过的坑:空闲降帧与 Pet 窗口帧率总闸、Live2D SDK 能改到哪一层、VRM 加载竞态、形象周边的弹窗和点击穿透、内置模型与动作文件处理。改 static/live2d、static/vrm、static/mmd、static/avatar、static/frame-pacing.js、static/subtitle/subtitle-window.js,或者要讨论「渲染性能还能怎么优化」之前读。
92%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
The skill exposes the agent to untrusted, user-generated content from public third-party sources, creating a risk of indirect prompt injection. This includes browsing arbitrary URLs, reading social media posts or forum comments, and analyzing content from unknown websites.
The skill is a maintainer guide for the avatar rendering code, and its only outsider-adjacent text is review-comment triage (PR review feedback judged as 收/不收/驳回). That is GitHub PR text, but the agent reads it only after a specific PR or review is selected, not from a feed it monitors, so it is low rather than medium.
Project-N-E-K-O/N.E.K.O
e14f619
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.