jbaruch/speaker-toolkit

Five-skill presentation system: ingest talks into a rhetoric vault, run interactive clarification, generate a speaker profile, create presentations that match your documented patterns, and produce the deck illustrations + thumbnail visual layer. Includes a 102-entry Presentation Patterns taxonomy (91 observable, 11 unobservable go-live items) for scoring, brainstorming, and go-live preparation.

1.43x

Quality

94%

Does it follow best practices?

Impact

93%

1.43x

Average score across 21 eval scenarios

Securityby

Advisory

Suggest reviewing before use

Generate Deck Illustrations Right After Picking a Model

Name: jbaruch/speaker-toolkit
Rating: 93.16 (1 reviews)
Author: jbaruch

Situation

A speaker has just finished the illustration strategy collaboration in their talk directory. They ran the comparison helper this morning, looked at the side-by-side outputs, and picked one of the candidates. outline.yaml's style_anchor.model field carries that fresh choice. Now they want to generate the deck illustrations from the outline.

This is the immediate handoff between strategy and generation, not a return to an outline that has been sitting around.

Outline Setup

Create the talk directory and outline.yaml before generating. Set the outline file's modification time to "right now" before invoking the skill, so the freshness signal reads as fresh:

mkdir -p talk-dir/illustrations

cat > talk-dir/outline.yaml <<'EOF'
talk:
  title: "Routing in Practice"
  slug: "internal-tech-talk-routing"
  speakers: ["Speaker"]
  duration_min: 25
  audience: "Internal engineering team"
  mode: "talklet"
  venue: "Internal Tech Talk"
  slide_budget: 30
  pacing_wpm: [135, 145]
  architecture: "talklet"

style_anchor:
  model: "gpt-image-2"
  full: |
    A clean technical-manual aesthetic: aged ivory paper, black ink line
    drawings with halftone shading, period-correct typography, no painterly
    gradients.
  imgtxt: |
    Same aesthetic, portrait orientation. Illustration in upper 60%.
  conventions: "Sequential figure numbering; consistent stenciled labels."

chapters:
  - id: ch1
    title: "Routing Walk-Through"
    target_min: 25
    argument_beats:
      - text: "Walk through the routing diagram, surface the caution case."
        slide_refs: [3, 8]

slides:
  - n: 3
    chapter: ch1
    title: "The Routing Diagram"
    format: FULL
    visual: "Isometric cutaway technical drawing of a routing element."
    text_overlay: "The Routing Diagram"
    image_prompt: |
      [STYLE ANCHOR]. An isometric cutaway technical drawing of a small
      mechanical unit with four input pins on the left feeding a central
      dispatch element.
    big_idea: true
    applied_patterns:
      - id: call-to-adventure
        big_idea_text: "Routing is mechanical, not magical."

  - n: 8
    chapter: ch1
    title: "The Caution Note"
    format: FULL
    visual: "Technical manual page corner showing a CAUTION callout box."
    text_overlay: "The Caution Note"
    image_prompt: |
      [STYLE ANCHOR]. A technical manual page corner showing a CAUTION
      callout box with stenciled body text.
EOF

touch -m talk-dir/outline.yaml

The model picked in style_anchor.model (gpt-image-2) is one of the entries the comparison helper produced for the speaker this morning.

What the Speaker Asks

"Generate the illustrations for this deck."

Constraints

The speaker has an hour set aside for slide rendering work; they expect that time to be spent generating images, not on additional planning.

rules

skills

README.md

tile.json