visual-debugging
Owns how a picture is judged and corrected, whatever made it, namely the instrument, planned captures of matched stills, passes, comparisons and correction rounds. Use before judging a form, a motion, a guide pass, a generated image or a film on screen, before claiming a visual change works, and when changing viewer or capture behavior or finding it broken or short of a view the judgment needs.
- License 许可
- MIT
- 最近更新
- 2026-11-15
- Stars
- ★ 15
- Forks
- 3
01 前言 Front Matter
- name
- visual-debugging
- description
- Owns how a picture is judged and corrected, whatever made it, namely the instrument, planned captures of matched stills, passes, comparisons and correction rounds. Use before judging a form, a motion, a guide pass, a generated image or a film on screen, before claiming a visual change works, and when changing viewer or capture behavior or finding it broken or short of a view the judgment needs.
02 资源文件 6
- 参考文档capture.md5.3 KB
- 参考文档comparison.md2.3 KB
- 参考文档correction-rounds.md2.1 KB
- 参考文档instrument.md9.7 KB
- 参考文档passes.md2.1 KB
- 参考文档rendering-cost.md520 B
03 SKILL.md 正文
Visual debugging
Readiness or numerical success proves no appearance; only an observed picture does. This skill owns how a picture is captured, compared and corrected, whatever made it: a form, a motion, a guide pass, a generated image or an encoded film. The skill of each subject owns what to look for in it.
An agent reads a picture only as stills, so each judgment starts from a planned capture. The instrument that plans and draws those stills is the inspection and capture core of render, which the inspector page of viewer and the capture host of cli both drive with one request text.
| Task | Read |
|---|---|
| Knowing what the instrument draws, plans, records and leaves out | Instrument |
| Restoring the instrument, or planning a capture's subject, views, times and stills | Capture |
| Choosing what each still draws, and marking what a measurement found | Passes |
| Setting two captures against each other | Comparison |
| Correcting what a capture shows | Correction rounds |
| Measuring what drawing costs | Rendering cost |
| Subject | Read also |
|---|---|
| A form, its construction, rig or materials | Inspecting a form |
| Body motion | Motion verification |
| The Human figure | Human verification |
The package boundaries say which package owns display, capture and each measurement. Add a view, pass or selection the instrument lacks at that owner under development, so every consumer of the instrument gains it.
04 同仓库更多 Skills
story-workflow
Chooses the story route and who writes the screenplay, and routes to each route's steps and to completion. Use when starting, resuming or completing narrative work.
story-correction
Inspects a narrative body and repairs it by cause, at its earliest owner and within closed boundaries. Use before evidence, after review findings, or when requirements or scope change.
development
Implements and tests production source, integrations, dependencies and commands, reusing installed packages before writing a capability. Use for code or host changes and behavioral tests; documentation owns technical writing and story owns creative bodies.
pull-request
Branches, commits, pushes, opens, updates, and merges automovie pull requests. Use only when the user asks for one of those actions or a mandate names delivery.
issue-campaign
Runs automovie's issue campaign, repeating full discovery rounds until one finds nothing new, publishing the issues, and landing them in one pull request per cycle. Use for a broad audit or repeated issue-to-pull-request work, not for one defined issue.
documentation
Owns how any document is written, with its sentences, terms, paragraphs and lists, and what a README holds. Use before writing or changing a technical document, README or doc comment; instructions for an agent add the prompting skill's rules.