使用方法
- 复制安装命令，在自己的智能体环境安装前先检查源码包。
- 从原始文档核对所需工具、凭据和支持的宿主。
- 说明输入材料与期望交付物，并对照来源检查输出。
npx skills add https://github.com/vasanthsreeram/alvarmethod --skill learn-visual开始前需了解
- 安装不包含模型订阅、API 额度或访问文件的权限。
- 仓库星数和安装量反映活跃情况，不等于独立质量或安全测试。
仓库原始内容
按源仓库内容呈现，保留标题、案例、代码、表格、链接以及原文引用的演示图片。
Learn visual
One idea, one picture. The picture exists so the learner can accept the step — not as decoration.
Output
Write .alvar/visuals/<slug>-<n>.svg (create the folder). Embed or link it in the session file.
Prefer SVG. Use another format only if the harness cannot preview SVG.
Loop (do not skip)
- State the claim the picture must make, in one sentence.
- Draw the smallest picture that makes that claim.
- Look at the file (image/read tool). If you cannot view it, say so and keep the SVG simple enough to audit as text.
- Fix labels, overlap, wrong arrows, missing units, or a picture that does not match the claim.
- Look again. Stop after a clean pass, not after the first draft.
Design
- One claim. No collage of the whole course.
- Large labels. High contrast. No tiny legend the learner needs a second lesson to read.
- If the idea is algebraic, show the objects (arrows, planes, machines), not a screenshot of the equation.
- Do not add decorative gradients, watermarks, or "AI art" backgrounds.
Failures to catch on the look pass
- Arrow direction disagrees with the prose
- Two symbols for the same object
- 3D that hides the relation
- Cropped text
- A picture of a different special case than the one just taught
After
Tell the learner what to look at first ("the two arrows, then the pairing number"). Do not re-teach the whole node unless they ask.
