agentleFS
mikel-codemikel-code/claude-skills/deck-design-prompt/SKILL.md

deck-design-prompt

Skill0 starsChanged 3 months agoMITView on GitHub

Reads a completed deck.md (produced by mckinsey-deck) and outputs a ready-to-paste prompt for Claude Design that tells it how to apply visual design without touching any content.

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 6 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Very specific:

    Read deck.md in the current directory, or the path the user specifies.

    line 13
  2. Very specific:

    Every --- is a hard slide boundary — do not merge or split under any circumstances.

    line 32
  3. Distinctive:

    Never promote sources into the body, and never delete them.

    line 47
  4. Distinctive:

    Do not truncate action titles — they are deliberately long complete sentences.

    line 62
  5. Very specific:

    Do not merge or split slides — each --- separator is intentional

    line 63
  6. Distinctive:

    Do not add, remove, or reorder any content

    line 64
  7. Distinctive:

    Do not paraphrase or summarise any text

    line 65
  8. Distinctive:

    Do not add a table of contents, agenda slide, or any slide not present in the Markdown

    line 66
  9. Very specific:

    Do not render the <!-- comparison: ...

    line 67
  10. Distinctive:

    Do not fabricate any data, comparator, axis, or trend to satisfy a chart.

    line 68
  11. Very specific:

    After outputting the prompt, say nothing else unless one of these applies: a slide-count discrepancy, or one or more chart-suppressed / ⚠ …

    line 72

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Deck-to-Design Prompt Generator

Produce a Claude Design prompt from a finished deck.md file. The prompt tells Claude Design exactly how to interpret the Markdown structure and what it must never touch.

Steps

  1. Read deck.md in the current directory, or the path the user specifies.
  2. From the manifest comment (<!-- deck: <goal> | slides written: N of M -->), extract the deck goal and slide count.
  3. Count every --- separator in the file to confirm total slides (if the count differs from the manifest, use the actual count and note the discrepancy to the user after outputting the prompt).
  4. Collect all ## lines in order — these are the section-intro slides.
  5. Read each content slide's <!-- comparison: <type> --> tag. For any tag that is a real comparison (not none), verify the slide body actually holds a comparator — a second comparable series, prior period(s), a peer, a target, or a threshold. A lone single value is not a comparator.
  6. Build the no-chart list: every content slide whose tag is non-none but has no comparator, plus every slide whose body carries a ⚠ Missing context flag. This skill is read-only — never edit deck.md. Instead, the prompt will instruct Claude Design to render these slides as text/table, and you will warn the user about them after the prompt so they can fix the content in mckinsey-deck (add the comparator, or let its final pass downgrade the tag to none).
  7. Fill in the template below and output it. Output only the prompt — no preamble, no commentary, no code fences around it. The user will copy and paste it directly into Claude Design.

Output template


Convert the attached Markdown file into a professional presentation deck. Do not alter, summarise, or reorder any content — every word in the Markdown is deliberate. Your job is to apply visual design, not editorial judgment.

Context [Goal summary from the manifest comment — one sentence, plain text]

Format

  • One slide per --- separator. This deck contains [N] slides total.
  • Every --- is a hard slide boundary — do not merge or split under any circumstances.

Slide hierarchy

  • ## NN — Section title → Section-intro slide (divider). Display the short title only; no body content exists on these slides.
  • ### Full action sentence → Content slide. The ### heading is the slide title — display it in full. Everything below the heading is the slide body.

Sections in this deck [Numbered list of the ## lines in order, e.g.:

  1. 01 — Opening
  2. 02 — Situation …]

Footnotes & sources

  • Inline [N] markers next to a figure or claim are reference markers — render them as superscript markers on that figure/claim.
  • The Sources: block at the foot of a slide lists the full references. Render it as a single small, bottom-aligned source line (footnote zone), keeping the [N] numbering intact.
  • Never promote sources into the body, and never delete them.

Comparison tags & charts

  • A hidden <!-- comparison: <type> --> on a content slide is a chart-form directive — never render the comment as visible text. Build at most ONE exhibit per slide, and it must make the action title's point:
    • component → share view (pie or 100% stacked bar)
    • item → ranked bar
    • time-series → column (few periods) or line (many)
    • distribution → histogram
    • correlation → scatter or paired bars
    • none → no chart; render as text, a table, or a single callout figure
  • Carry the slide's source line on the exhibit; one message per exhibit (the action title is its takeaway); declutter — no 3D, no chartjunk.
  • Never invent data, a comparator, an axis, a trend, or a series not present in the slide body. If a slide lacks the data to build its indicated chart, render it faithfully as text/table instead of fabricating a chart.
  • Render these slides without a chart (text/table only): [list slide numbers/titles from the no-chart list, or "none"]

Don'ts

  • Do not truncate action titles — they are deliberately long complete sentences; truncating them destroys the meaning
  • Do not merge or split slides — each --- separator is intentional
  • Do not add, remove, or reorder any content
  • Do not paraphrase or summarise any text
  • Do not add a table of contents, agenda slide, or any slide not present in the Markdown
  • Do not render the <!-- comparison: ... --> comments as visible text — they are directives, not content
  • Do not fabricate any data, comparator, axis, or trend to satisfy a chart; fall back to text/table when the data is not on the slide

After outputting the prompt, say nothing else unless one of these applies: a slide-count discrepancy, or one or more chart-suppressed / ⚠ Missing context slides — in which case list them briefly so the user can add the missing comparator in mckinsey-deck.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.