Leadership Speech Generator

Problem it solves

Leadership speeches usually have to satisfy occasion, speaker identity, policy grounding, data and cases, structure, official document formatting, and Word delivery at the same time. Manually sourcing material, checking policy wording, and fixing layout is slow, and drafts can become vague, contain inaccurate citations, or use a structure that does not fit the meeting.
This skill turns the task into a constrained workflow. It uses parameter extraction, search records, an instruction archive, planning output, and checklist gates to reduce drift from ad-hoc writing.

Core capabilities and key steps

• Information collection: first confirm the speech title, region or organization, speaker role, extra requirements, and length.
• Source analysis: when files, notes, or knowledge base entries are provided, extract policy basis, key data, typical cases, distinctive wording, and outstanding issues.
• Parameter extraction: produce an 8-parameter JSON record covering the core theme, geographic scope, identity label, occasion type, time context, and length.
• Web research: run repeated searches across policy background, industry trends, comparable speeches, data support, and key phrasing, then retrieve high-value original sources.
• Instruction archive: generate a 13-field archive, select a structure such as the classic three-part model, task-deployment model, or problem-driven model, and map material to content modules.
• Drafting and review: complete an 8-dimension planning pass before writing, then draft the speech and verify formatting, policy citations, structural proportion, speaker identity, and the .md / .docx deliverables.

Boundaries and notes

• Use it for Chinese leadership speeches or formal meeting remarks, not notices, requests for instructions, news articles, academic papers, or purely personal speeches.
• The skill enforces hard constraints on search volume, archive fields, and self-check items. Delivery may be incomplete if web access or Word generation dependencies are unavailable.
• Policy document numbers, speaker names, local cases, and statistics should still be manually verified to avoid misattribution.