leo-skills

Use when the user needs to modify a PDF template — replacing variables like names, dates, amounts, or branding elements while preserving the original layout, fonts, colors, underlines, and spacing. Covers the full pipeline: analyze PDF structure (text layer vs image layer), OCR text blocks, classify each block (static label vs fill-in variable), detect underlines and mixed styles within lines, extract font sizes / colors / weights / indentation levels, build a structured doc_definition.json, and render a new PDF via a reusable engine. Applies to any image-layer-dominant PDF (offer letters, contracts, certificates, invoices, forms).

Use when starting a new feature or project, need to establish a spec→plan→task→code workflow, want AI to follow documentation-before-code discipline, or need to set up automated quality gates (Harness) for AI-generated code. Also use when AGENTS.md is growing too large, AI is drifting from spec, or you need to integrate BDD acceptance criteria into your development process. Applies to React frontend + Python/Go backend stacks.

Use when building a ChatBI (conversational BI) MVP from scratch, need to understand the 5 core capabilities (NL2SQL, multi-turn dialogue, RAG knowledge base, data visualization, intelligent attribution), or want to reference SuperSonic's architecture to guide implementation

生成地道英语口语"句模"（Sentence Model）——围绕一个话题提供6个高频提问+12句母语级长句，标注词伙、句型、观点，支持拆解复用、跨话题联动和复述法练习。五大能力：(1)生成句模——"帮我生成一个关于旅行的句模""来套XX话题的口语素材"；(2)拆解复用——"拆解一下""提取里面的词伙和句型"；(3)跨话题联动——"能把食物的句型用到科技话题吗"；(4)轻量练习——"帮我练练""出个题考考我"；(5)复述法实操——"帮我复述这句""用复述法练一下""用自己的话说一遍""retell this"。当用户提到"句模""口语话题""口语素材""英语口语""sentence model""oral English""复述""retelling""口语练习方法""口语怎么练"，或需要系统性准备口语素材、拆解句子提取词伙句型、跨话题复用、用复述法内化口语表达时，使用此skill。