ワンクリックで
submit-feedback
Read a UI feedback markdown file exported by designer-notes and apply targeted code revisions
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Read a UI feedback markdown file exported by designer-notes and apply targeted code revisions
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Read a UI feedback markdown file exported by designer-notes and apply targeted code revisions
Wire designer-notes into the current project — adds the script tag, starts the dev server, and opens in browser for Figma-style feedback commenting
Wire designer-notes into the current project — adds the script tag, starts the dev server, and opens in browser for Figma-style feedback commenting
| name | submit-feedback |
| description | Read a UI feedback markdown file exported by designer-notes and apply targeted code revisions |
Read structured UI feedback exported by designer-notes.js and make targeted revisions to the source code.
Locate the feedback file. If a path argument is provided, use it. Otherwise:
feedback-*.md files~/Downloads/ for feedback-*.md files, sorted by most recentRead and parse the feedback file. The file is grouped by page. Extract from each ### Comment N block:
**Element:** line — content inside backticks)**Tag:** line)**Position:** line)> lines)## Page: header)Identify the source files. Map each ## Page: section to the actual HTML file in the project.
For each comment, locate the code:
Present a summary table:
| # | Page | Element | Feedback | Proposed Action |
|---|---|---|---|---|
| 1 | test.html | h1 "Welcome..." | Too generic | Revise heading text |
Ask how to proceed. First check dn-config.json for preferences.autoApply. If true, skip this step and apply all comments automatically. Otherwise ask:
Check for repeat comments. Before applying changes, read feedback-archive.md (if it exists) and check whether any current comments target the same CSS selector as a previously archived comment. If a match is found:
Apply changes for each comment:
#o3, #gpt-4o, #o4-mini): These indicate the user's preferred model for this comment. If Codex supports model switching mid-session (e.g., via /model command or multi-agent dispatch), apply it. Otherwise, note the preference but process with the current model. Strip directives from comment text.#high-effort, #medium-effort, #low-effort): Adjust your reasoning depth accordingly. Strip from comment text.#opus, #sonnet, #haiku. Strip these — they are for Claude Code./keyword references (e.g., /arrange, /polish, /bolder). These are design intent hints — interpret them as direction for how to approach the edit. Strip the /keyword from comment text.After all changes, present an action log:
| # | Comment | What Changed | File:Line |
|---|---|---|---|
| 1 | "add a line break" | Added <br> before heading | test.html:482 |
Write changelog HTML. Update .designer-notes/changelog.html. Read ~/.agents/skills/designer-notes/changelog-template.html for the HTML shell. If it exists, replace <!-- SECTIONS --> with section elements. Prepend new sections before existing ones.
Archive feedback:
curl -s -X POST http://localhost:3847/archive-feedback
Let the user know changes are live and the browser will reset automatically.
Check for updates. Run the update checker:
node ~/.agents/skills/designer-notes/check-update.js
If it produces JSON output, mention at the end of your message: designer-notes v{latest} is available (you have v{installed}). Run {command} to update. If no output, skip this — the user is up to date. Only run this check once per session — if you've already shown an update nudge in this conversation, skip it on subsequent runs.
(Same format as Claude Code — generated by designer-notes.js)
# UI Feedback
Generated: [timestamp]
Total comments: [N]
Viewport: [width]x[height]
---
## Page: [page-path]
### Comment 1
**Element:** `[CSS selector]`
**Tag:** [TAG] | **Text:** "[preview]"
**Position:** click at (x, y) on page; offset (x, y) within element
**Element bounds:** WxH at (x, y)
> [The actual feedback text]
@media query