agentleFS
samulee003samulee003/agentic-ppt-skills/.agents/skills/presentation-prototype/SKILL.md

presentation-prototype

Skill0 starsChanged 24 days agoMITView on GitHub

Builds a representative prototype in the real slide source so the team can test tone, density, visual system, and page grammar before producing the full deck. Use when the deck-prototype gate is current for a presentation project or when a router asks for representative opening, thesis, evidence, dense-page, and ending pages plus live review in the chosen slide engine.

The file

Show its rules and text

We split this skill into the rules it gives an agent

14 rules. About 0 are common, 5 are distinctive, and 9 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 9

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use when deck-prototype is the current gate.

    line 8
  2. Very specific:

    If PRESENTATION-BRIEF.md (or the gate-1 intake) already records the engine, use it.

    line 14
  3. Very specific:

    Read .agents/adapters/<engine>/README.md first for that engine's prerequisites and setup, then .agents/adapters/<engine>/prototype.md for …

    line 17
  4. Very specific:

    Read repo instructions, PRESENTATION-BRIEF.md, EVIDENCE-LEDGER.md, STORY.md, STORYBOARD.md, cited source files, and existing slide files …

    line 22
  5. Distinctive:

    Always explore the files instead of asking the user for discoverable facts.

    line 23
  6. Very specific:

    Follow the engine's authoring rules (from .agents/adapters/<engine>/prototype.md) before editing any page in the real slide source.

    line 25
  7. Distinctive:

    Use real assets, real type scale, and the intended visual direction rather than placeholder visuals or compressed mini-layouts.

    line 27
  8. Very specific:

    Reject AI 3-tier text walls: Do not generate Title + Subtitle + 3-4 dense bullet paragraphs.

    line 34
  9. Distinctive:

    Balanced eyeline centering: Keep visual weight centered along the optical middle (y: 35%–65%).

    line 35
  10. Distinctive:

    Split over shrink: When space is tight, split the concept across multiple slides rather than shrinking fonts below projector-safe …

    line 39
  11. Very specific:

    Arena (optional): when the visual direction is unsettled, the deck is high stakes, or the first prototype was rejected twice, follow …

    line 51
  12. Very specific:

    Unslop: run node .agents/skills/make-presentation/scripts/unslop-check.mjs <slide-folder> on the prototype pages and fix findings per …

    line 52
  13. Distinctive:

    Keep the prototype in the real slide source so later work can revise the actual pages instead of rebuilding from mockups.

    line 53
  14. Very specific:

    Write only consequential decisions to GRILL-LOG.md from ../presentation-product-grill/templates/GRILL-LOG.md.

    line 57

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Build the presentation prototype

Use when deck-prototype is the current gate.

Engine

This gate renders slides, so it needs a slide engine. Resolve it before authoring:

  • If PRESENTATION-BRIEF.md (or the gate-1 intake) already records the engine, use it.
  • Otherwise ask the user, one question, with a recommended default. Offer installed engines first (see .agents/adapters/: open-slide, genoffice, marp, pptxgenjs); the default reference adapter is open-slide (or genoffice when the user wants a native byte-preserving Office .pptx via GenOffice).
  • Verify the adapter exists before recording the engine: run test -d .agents/adapters/<engine> (or ls .agents/adapters/ to list installed ones). If it fails, tell the user there is no adapter for that engine and offer the installed ones — do not record an engine with no adapter, and do not invent the adapter's rules inline.
  • Read .agents/adapters/<engine>/README.md first for that engine's prerequisites and setup, then .agents/adapters/<engine>/prototype.md for its authoring rules, canvas contract, and live-review surface. Author to that adapter — do not invent engine-specific rules here.
  • Record the choice so downstream gates (stage-grill, production-qa) inherit it: set it in the status tracker via node .agents/skills/make-presentation/scripts/status.mjs set-engine <work-root> <engine>, and add an ## Engine line to PRESENTATION-BRIEF.md.

Core design principles & anti-pitfall rules

  • Read repo instructions, PRESENTATION-BRIEF.md, EVIDENCE-LEDGER.md, STORY.md, STORYBOARD.md, cited source files, and existing slide files before editing.
  • Always explore the files instead of asking the user for discoverable facts.
  • Grill in rounds per ../make-presentation/references/grilling-rounds.md when consequential prototype decisions are unresolved: ask the whole frontier in one round, each question with a recommended answer worded so "yes" accepts it.
  • Follow the engine's authoring rules (from .agents/adapters/<engine>/prototype.md) before editing any page in the real slide source.
  • Flexible Deck Length (No Arbitrary Page Limit): The final presentation length is unconstrained by arbitrary page count caps (e.g. not fixed to 18 slides). The total slide count scales organically with lecture duration, audience interaction, and narrative depth. At the prototype gate, build the representative set in references/prototype-set.md (or the full draft if appropriate) to prove the visual language and page grammar.
  • Use real assets, real type scale, and the intended visual direction rather than placeholder visuals or compressed mini-layouts.
  • Author to the engine's canonical projector-safe typography and overflow rules. Do not create a separate type scale in this workflow.
  • Confirm one necessary job and one dominant message per slide.
  • Reject verbose paragraphs and nested text tiers.
  • Split the slide rather than shrink text or hide overflow.

Aesthetic guardrails (learned from production)

  1. Reject AI 3-tier text walls: Do not generate Title + Subtitle + 3-4 dense bullet paragraphs. Maximize Signal-to-Noise Ratio (SNR). One dominant message per slide.
  2. Balanced eyeline centering: Keep visual weight centered along the optical middle (y: 35%–65%). Never collapse all content or concluding punchlines down to the bottom border. Cover/title slides must boldly command the entire canvas.
  3. High-craft iconography: Strictly forbid OS emojis (🚀, 💡, 🔥, 👶) and single-character boxed glyphs ("寫", "參"). Use vector Flat SVG icons or dedicated custom illustrations.
  4. Integrated visual metaphors: In complex diagrams (e.g. iceberg, funnel, 3-pillar framework), embed callouts directly into the visual structure; do not resort to lazy 50/50 side-by-side splits or crop critical diagram landmarks.
  5. Balanced multi-item grids: For 5 items, use 3-top / 2-bottom (or 2-top / 3-bottom), never a vertical stack of 5 narrow stripes.
  6. Split over shrink: When space is tight, split the concept across multiple slides rather than shrinking fonts below projector-safe thresholds.
  7. Instructional interaction layouts: Provide clean visual templates for active audience engagement (e.g., audience polling/temperature cards, buzz-pair discussion prompts, step-by-step dialogue flows, and takeaway scaffolding).
  8. Language discipline & purity:
    • Guarantee 100% target language consistency on-screen (zero stray untranslated characters).
    • For Chinese presentations with oral delivery: keep on-screen slide copy in concise standard written form, while keeping spoken vernacular (e.g. Cantonese oral script) in the Speaker Notes.
  9. 4 Page Archetypes (Prohibit Uniform 2-Box Monotony): Vary slide composition intentionally across four distinct archetypes:
    • Archetype 1 — Big Typography / Hero Statement (大字流): Dominant display typography or hero number with zero bullet lists for thesis statements, golden rules, and transitions.
    • Archetype 2 — Split Visual + Action Grid (半圖半文): Asymmetric 2-column layout pairing a framed visual/vignette on one side with structured action steps on the other.
    • Archetype 3 — Full-Canvas System Metaphor (大圖流): 70%+ of the content zone dedicated to a structured diagram or visual model with callouts embedded inside the nodes.
    • Archetype 4 — Distinct Multi-Card Grid (卡片流): 2–4 high-craft cards with committed borders/shadows and clear visual hierarchy (avoid generic 1px faint gray SaaS dashboard boxes).
  • Arena (optional): when the visual direction is unsettled, the deck is high stakes, or the first prototype was rejected twice, follow ../make-presentation/references/arena-prototype.md to build 2 to 3 competing directions, pick a base, and graft. Skip in fast-track mode.
  • Unslop: run node .agents/skills/make-presentation/scripts/unslop-check.mjs <slide-folder> on the prototype pages and fix findings per ../make-presentation/references/unslop-copy.md.
  • Keep the prototype in the real slide source so later work can revise the actual pages instead of rebuilding from mockups.
  • Require a live review in the chosen engine before calling the gate passed.
  • Return to story or evidence if representative pages fail because the narrative job, proof, or density is wrong rather than merely unpolished.

Write only consequential decisions to GRILL-LOG.md from ../presentation-product-grill/templates/GRILL-LOG.md.

  • Append new decisions without deleting or rewriting entries from other grill phases or earlier runs.
  • Preserve existing decisions when revisiting the prototype gate.
  • Log changes that materially affect visual system, page grammar, asset direction, representative scope, or what later full-deck work must preserve.

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.