On-site Expression Cue Cards
Cue cards are the cognitive interface for live presentations, not a lecture script with reduced font size. This skill first organizes the presentation task, then uses an original local tool to generate traceable card sets. It does not perform scientific review, nor does it output Word documents.
Card Syntax
Each card first clarifies five things, which are then mapped to JSON fields:
- Trigger Point: When to flip to this card. Written in
titleor beginning of the body. - Core Speech: Content that must be clearly stated live. Written in
body. - Evidence Anchor: Sources, data, or boundaries to verify. Written in
source. - Delivery Action: Pause, question, demonstration, or handover cues. Written in
notes. - Holder: One or more real roles. Written in
speaker. Do not guess responsibilities based on name or gender.
For question cards, the expected key points can be placed in the teacher version's notes; when creating the student version, remove answers from the input copy. notes are printed and are not hidden annotations.
Generation Workflow
- First confirm the scenario, speaker, card size, usage distance, and output format. Scientific conclusions, experimental data, competition endorsements, and personal endorsements must be verified before card generation.
- Build the structure from the blank template, fill it according to the operation guide. The teaching simulation example serves only as a format reference.
- Run
scripts/cue_cards.py. HTML mode uses only Python standard library; PDF requires ReportLab and a user-provided TrueType font that is embeddable and covers required Chinese characters. - Check the complete set, each speaker's subset, shared card assignments, card numbers, continuation card order, full source index, and whether all body characters are present.
- Verify PDF page-by-page rendering; use the target browser for print preview of HTML. Record the medium that has been verified. Do not equate on-screen viewing with actual print acceptance.
Typographic Contract
- Default: one card per page, 150×100mm. Body text default 13pt. Size and font size can be adjusted via command line.
- Long content is preferentially continued at sentence or line breaks. If it cannot fit completely, continue splitting without silently deleting characters.
- The complete set and subsets share stable card numbers. Cards shared by multiple people go into each holder's subset.
- Each page displays sources. Overly long sources point to the
index.htmlfile included in the package, with full values retained in both the index JSON and HTML. - Errors are returned for missing glyphs in the font, unplaceable titles, invalid JSON, or existing output directory. The program does not actively delete or overwrite previous output.
The input model does not store formulas, scientific charts, images, or rich text. If the task requires faithful conversion of such content, use document or presentation tools instead. This tool only reads local JSON, does not access URLs within it, and does not execute commands in the body.
For first-time use, read the introduction. Rehearsal and evaluation methods are found in the teaching guide. When handing over to a course platform or video module, keep stable card numbers and source anchors so that live expressions can be traced back to the original content.