Xiao Zhu AI Performance Review PPT Generator Codex Edition

Problem

Workplace reviews, promotions, probation confirmations, and sensitive management updates often fail not because of too few slides, but because evidence is scattered across chats, project files, meeting notes, and old decks. Facts, inferences, and personal requests get mixed together, producing slogan pages, crowded pages, or complaints about mechanisms.

How It Works

The skill turns existing materials into a traceable reporting chain: materials -> facts -> problem -> method -> outcome -> value -> next steps. It starts with requirements and a template contract, asking only questions that materially change the result. Then it builds a PPT strategic SPEC with Problem Statement, Audience, Desired Judgment, and Evidence Base; after confirmation, it generates an outline where each slide answers “what should the leader understand?” Page design enforces one management judgment per slide, leads with the conclusion, separates facts from inferences, and routes weak evidence into a data-gap list. Before delivery, it renders and checks every page for alignment, spacing, overflow, missing assets, and template damage. Deliverables typically include Speaker Notes, an asset source list, pending data list, and page-by-page visual validation.

Boundaries

It fits scenarios with company templates, fixed sections, frozen pages, or a clear decision objective, especially reviews, promotions, probation confirmations, and sensitive management reporting. If the user wants quick templating, invented polished pages, or copied layouts from another deck into the company template, the skill will not generate directly; it prioritizes confirmed evidence and template constraints.