scholarflow-research-slides

Use when a researcher needs one self-contained presentation slide designed, drafted, or revised as editable HTML with a documented visual specification.

跳到安装

仓库: zhoy0409-debug/scholarflow-skills
最近来源活动: 2026年7月19日 07:08
检测到的 SKILL.md 语言: 英语
星标: 0
分支: 0

安装方式

默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

直接命令不会经过审查 Prompt；运行前请先检查来源。

npx skills add https://github.com/zhoy0409-debug/scholarflow-skills --skill scholarflow-research-slides

命令会保持在同一行。复制前请横向滚动并检查完整内容。

想先保存到本地？可下载 SkillsMP 当前能够提供的文件。

检查来源文件

决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。

文件资源管理器

7 个文件

SKILL.md

来源说明 · 只读预览

返回搜索返回安装

polish-sci-figures

Use when a researcher needs scientific figures created, redrawn, arranged, audited, or exported with readable typography, high-contrast color, non-overlapping legends, and publication-ready files.

2026-07-19

bioinformatics-workbench

Use when a researcher needs to turn a biological question and available sequencing or omics data into a realistic, reproducible bioinformatics workflow with explicit QC gates and deliverables.

2026-07-19

journal-selection-advisor

Use when a researcher needs journal selection, manuscript-journal fit analysis, tiered recommendations, JCR or CAS target matching, institutional requirement checks, publication-risk calibration, or submission sequencing.

2026-07-19

查看这个仓库的全部 Skills

name	scholarflow-research-slides
description	Use when a researcher needs one self-contained presentation slide designed, drafted, or revised as editable HTML with a documented visual specification.

Single-page PPT authoring

Given:

A target slide directory (typically .pptwork/<deck>/<slide-name>/)
One outline entry (pageType, assetId, key content, source, transition)
The deck exportMode (raster or editable)

Produce:

design.md — frontmatter title + layout, then ## Content / ## Note / ## Design (all three sections must be non-empty)
slide.html — single-file self-contained, with the ## Note text mirrored into a <script type="application/json" id="ppt-speaker-notes-json"> island

Out of scope: cross-page planning (use the ppt skill's story-planning phase), screenshots and visual QC (use the ppt skill's screenshot.mjs + visual-qc reference), exporting (use export.mjs), web search (use the host app's webfetch / websearch).

File contract

`design.md`

frontmatter:
- title (required)
- layout (required, short English identifier matching the layout you used)
body, three sections, all non-empty:
- ## Content — page copy in human language: a conclusion-form complete-sentence title plus supporting bullets / short paragraphs.
- ## Note — speaker notes in natural prose (not bullets).
- ## Design — visual / engineering notes: layout choice rationale, transition from previous page, data source, TBD markers.
No empty placeholders. "This is an introduction page" is not acceptable.
Detailed schema → references/design-md-spec.md

`slide.html`

Must be single-file self-contained: CSS in <style>, JS in <script>.
Images: inline base64 or https CDN. Local relative paths (<img src="assets/...">, <link href="...">) are forbidden — they break in <iframe srcdoc> previews and in the static-server-based renderer unless the sidecar files happen to land next to the HTML.
Fonts, chart.js, d3 etc.: load from https CDN.
Mirror ## Note content into a <script type="application/json" id="ppt-speaker-notes-json">…</script> node. This is the only way speaker notes land in the exported PPTX.
For the slide to render at the correct PowerPoint canvas size, set html, body { width: 1280px; height: 720px; overflow: hidden; }.
Hard rules and common pitfalls → references/slide-html-rules.md (must read).
If exportMode=editable, also follow references/editable-html-rules.md.

Workflow

1. Read context

Outline entry (pageType / assetId / key content / source / transition) + the existing slide skeleton (design.md is an empty template, slide.html may only have <html> head/tail) + exportMode.

2. Look up the layout (assets contract)

If the outline names an assetId, look it up under the sister ppt skill's ../ppt/assets/<bucket>/index.json and read the matching descriptor:

Read	For
`index.json` entry's `htmlPath`	Find the HTML reference path
`index.json` entry's `materialDir` / `zones` / `authoringHints`	Understand the layout's slot composition
Resolved `html/<id>.html`	Borrow composition (version safe area, columns, hierarchy)
`specs/<id>.json` `zones`	Get structured per-slot descriptions (optional)

This skill ships no built-in template buckets — you (or the project) provide them. If no bucket is configured, skip this step and write slide.html from scratch using references/slide-html-rules.md and references/ppt-best-practices.md as your composition guide.

Do not pick the assetId yourself — that's a story-planning decision, the outline already names it (or explicitly says "no preset, design from scratch").

3. Don't copy the template HTML wholesale

Borrow only the composition (version safe area, columns, hierarchy, slot positions). Replace all placeholder copy ("Acme Corp", "2024 Q3 revenue", "scene illustration", etc.) with the page's real content. Placeholder bleed into a finished deck is one of the most common authoring failures.

4. Write `design.md` first

In the order Content → Note → Design:

## Content: conclusion-form complete-sentence title (not a noun phrase) + supporting bullets / short paragraphs.
## Note: speaker notes, natural prose, not bullets.
## Design: layout choice rationale, transition cue, data source, TBD markers.

5. Then write `slide.html`

Drive the HTML from the Content section. Whenever the data fits, prefer visuals over plain text:

Trends / proportions / comparisons → Chart.js (see references/diagram-and-chart.md).
Flow / architecture / system boundary → inline SVG.
Decorative imagery → base64 inline; for large images (>2MB) compress or use a remote CDN.
Icons → use a remote https SVG icon library (Heroicons, Lucide, Feather, Tabler, etc.) or paste a single inline SVG.

If exportMode=editable, write the HTML as PowerPoint-native objects: real text nodes for important copy, simple solid-fill shapes for cards, real tables, extractable Chart.js data, and separated background decoration. Do not place important text inside SVG/canvas/images/pseudo-elements.

6. Sync `Note` into the HTML

Mirror the ## Note text into the <script type="application/json" id="ppt-speaker-notes-json">…</script> island. This is the only path for speaker notes to land in the exported PPTX.

7. Need an external fact?

If a single page is missing one number / fact / citation / competitor data point:

Use the host app's built-in webfetch / websearch tool. This skill ships no web tool of its own.
Append findings to .pptwork/<deck>/materials/research.md (so they don't pollute the main context across slides).
Cite the fact in ## Design as source: research.md §<timestamp>.
Do not fabricate numbers.

8. Self-check

Any <img src="assets/..."> / <link href="..."> relative paths? Convert to base64 or remote https.
Note section synced into the HTML?
data-id / data-role attributes stable enough to support later edit / patch style touch-ups?

References (read on demand)

Doing	Read
Writing design.md	`references/design-md-spec.md`
Writing / editing slide.html	`references/slide-html-rules.md` (required)
Writing / rewriting editable deck HTML	`references/editable-html-rules.md`
Adding charts, inline SVG, diagrams	`references/diagram-and-chart.md`
Typography, font sizes, alignment, visual focus	`references/ppt-best-practices.md`

Decision rules

User says "tweak a sentence" → use edit / patch to mutate the matching data-id node, keep the outer structure intact.
User says "rework the layout" → rewrite the whole slide.html, but keep design.md's narrative intact.
Existing raster / complex HTML deck now needs editable output → rewrite each page with a language model according to editable-html-rules.md; do not rely on a pure rule-based converter.
Content is fundamentally about ratio / trend / comparison → lead with a chart, don't list raw numbers.
Content is fundamentally about flow / architecture / boundary → lead with inline SVG, don't pile up paragraphs.
A background / brand image is required → base64 inline; for large images (>2MB) compress first or use a remote CDN.
Data is missing → use the host's webfetch/websearch, do not invent numbers.

Common pitfalls

Leaving template placeholder copy in the finished page ("Acme Corp Case" in an unrelated deck).
Data page that's just "numbers + a paragraph" with no chart.
image-slot left with a "scene illustration" placeholder — the thumbnail will look empty.
A relative image source for a file that is not packaged - the iframe preview breaks.
Forgot to mirror Note into #ppt-speaker-notes-json — exported PPTX has empty speaker notes.
Title written as a noun phrase ("AI Agent market") instead of a conclusion sentence.
Faked a number instead of running a quick fetch / asking the user.

scholarflow-research-slides

来源信息

安装方式

检查来源文件

同仓库更多 Skills

同仓库更多 Skills

Single-page PPT authoring

File contract

design.md

slide.html

Workflow

1. Read context

2. Look up the layout (assets contract)

3. Don't copy the template HTML wholesale

4. Write design.md first

5. Then write slide.html

6. Sync Note into the HTML

7. Need an external fact?

8. Self-check

References (read on demand)

Decision rules

Common pitfalls

Single-page PPT authoring

File contract

design.md

slide.html

Workflow

1. Read context

2. Look up the layout (assets contract)

3. Don't copy the template HTML wholesale

4. Write design.md first

5. Then write slide.html

6. Sync Note into the HTML

7. Need an external fact?

8. Self-check

References (read on demand)

Decision rules

Common pitfalls

`design.md`

`slide.html`

4. Write `design.md` first

5. Then write `slide.html`

6. Sync `Note` into the HTML

`design.md`

`slide.html`

4. Write `design.md` first

5. Then write `slide.html`

6. Sync `Note` into the HTML