Report deck zh

report-deck-zh — Chinese briefing slides have a grammar; speak it

Chinese work briefings have a structural grammar — parallel titles, whole-part-whole, data-caption pages — that English slide tools don't speak. This skill encodes it.

Lineage (adaptation-line skill — read before judging this redundant)

Derived from / inspired by frontend-slides (Zara). Four questions:

1. Where does the original fail? frontend-slides builds HTML slides with English deck aesthetics. It does not know the conventions of formal Chinese reporting: titles must be parallel (对仗), the document must close the loop (总分总), data pages must carry caliber captions, hierarchy follows Chinese official-document progression, not free-form bullets.
2. Was the author a failed user of the original? Yes. The author has produced formal Chinese work briefings at volume; decks laid out with English-aesthetic tools read as structurally wrong in that setting — recognizably "not a formal document" before anyone reads a word.
3. What is the patch? Replace the English aesthetic templates with a Chinese-briefing structural grammar library: parallel-title generation rules, whole-part-whole skeleton, data-caliber page pattern, hierarchical progression rules. Visual tokens reuse the restraint system from read-not-decorate.
4. For whom is this a reason to switch? Anyone producing formal Chinese work briefings (述职 / 总结 / 立项 / 复盘) who needs structure that reads as native formal reporting, not an English pitch deck.

Stance (three load-bearing claims)

1. Formal Chinese reporting has its own structural grammar; English slide tools don't speak it. Mechanism: parallel titles, whole-part-whole, and caliber pages are conventions of the genre — apply English deck logic and the structure itself is wrong, regardless of polish. Field observation: a Chinese briefing laid out by an English-aesthetic tool shows non-parallel titles and no data-caliber page; experienced readers flag it as "not a formal document" at a glance. Boundary: verified on formal work briefings; untested on academic or government-publication formats.
2. Structure only, never content. Mechanism: the skeleton generalizes; the content is case-specific, and generating it both exceeds this skill's warrant and risks leaking real-world specifics. Field observation: the author's own production rules repeatedly draw this exact line. If input is insufficient to infer structure, stop and ask the briefing type — do not invent content to fill a skeleton.
3. Restrained visuals; reuse read-not-decorate tokens. Mechanism: a briefing must read professional, not decorated; the homogenized card-beautification of mass AI-PPT tools is precisely the failure mode. Field observation: homogenized beautification makes every briefing look identical; a restrained structural layout reads as more professional, not less. Boundary: this is a claim about formal briefings; light or playful decks legitimately want different visuals — out of scope here.

Core actions (do X → load Y)

Task	Load
Decide the deck skeleton (briefing type, sections, closing loop)	reference/structure.md
Generate or repair parallel titles	reference/titles.md
Lay out data pages with caliber captions	reference/data-pages.md
Apply visual tokens	reference/visual.md
See a complete worked example	examples/demo.html

Workflow gate (mandatory): before laying out any slides, report to the user (a) the briefing type you identified (述职 / 总结 / 立项 / 复盘 / other) and (b) the proposed section skeleton. Wait for confirmation. Layout without a confirmed skeleton is the most common failure.

Hard rules:

• Every title group at the same level must be parallel: aligned character count, symmetric grammatical structure, consistent verb-object form (reference/titles.md).
• Every page presenting figures carries a caliber caption: data source, statistical scope, time caliber (reference/data-pages.md). No caption, no data page.
• The deck closes the loop: opening states the whole, body parts it, closing restates the whole with the delta (reference/structure.md).
• No emoji, no gradients, no decorative icons, no stock imagery (reference/visual.md).
• Examples and placeholders are fictional generic-business scenarios only.

Six fields

• Version: 0.1 (2026-06). Changelog: initial release. Corrections will be marked in place, not deleted (reverse-marking policy).
• Retirement condition: when a bare model, unprompted, reliably produces parallel titles, whole-part-whole structure, and caliber captions for Chinese briefings, this skill retires.
• Failure behavior: if the input does not determine the briefing type or skeleton, stop and ask. Do not silently force a skeleton onto unclear input; do not generate content to compensate.
• Composition: composes-with read-not-decorate (shared visual tokens). Division of labor with frontend-slides: English decks → that skill; formal Chinese briefings → this one. Explicit user instructions override this skill.
• Time: judgments dated 2026-06. Chinese briefing conventions are stable; the AI-PPT tool ecosystem is not. Review in ~18 months.
• Boundary: verified on formal Chinese work-briefing layout. Not for English pitch decks, not for light/social slides, does not generate briefing content. Untested beyond these — treat as unverified there.