ray-writer

Write verified Chinese long-form articles and licensed translations from ideas, clippings, or drafts.

160|18|Updated Jul 13, 2026
One-click install
npx skills add https://github.com/imraywang/rayskills --skill ray-writer-imraywang
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ray-writer
Source: https://github.com/imraywang/rayskills/tree/main/skills/ray-writer
Command: npx skills add https://github.com/imraywang/rayskills --skill ray-writer-imraywang

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Turning raw ideas, clippings, or rough drafts into publishable Chinese long-form articles usually breaks down on fact accuracy, consistent voice, and platform formatting. This Skill runs a traceable writing pipeline that separates facts from opinions, enforces a personal voice, and validates output with a mechanical checker before delivery. ## Core Features & Use Cases - Original long-form writing: Builds a minimal writing task, a fact checklist, and a continuous single-author draft, then runs a natural-language self-revision and a four-layer quality gate covering truthfulness, article structure, emotional tension, and personal voice. - Translation and repost mode: Handles foreign-language or third-party articles through two mandatory gates—authorization (source_permission) and visible attribution—supporting full translation or digest modes without ghostwriting or content laundering. - Mechanical quality checks: The article_check.py script validates length, headings, bold usage, template phrases, blank lines, and translation attribution fields, returning errors and warnings as JSON. - Knowledge base integration: Files drafts, research, and published pieces into an Obsidian-compatible vault structure, and hands off to cover generation, WeChat formatting, and X Articles skills. - Use Case: You have an interview transcript about a founder's pivot. The Skill establishes the core conflict, verifies facts, writes a 3000–8000 character article with mobile-friendly headings and bolding, runs the checker, and delivers the draft with a fact boundary record. ## Quick Start Use ray-writer to turn this idea and these source materials into a verified Chinese long-form article for WeChat and X.

Frequently Asked Questions about ray-writer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write a Chinese long-form article from raw materials?▼

Provide your idea, clippings, or draft and the Skill builds a minimal writing task defining reader takeaway, core conflict, and fact boundaries. A single author writes the full draft continuously, followed by a natural-language self-revision and a mechanical check via article_check.py before delivery.

How do I translate a foreign article for WeChat or X?▼

Translation tasks enter a dedicated mode requiring two gates: authorization (source_permission must be granted or open-license) and visible attribution in both frontmatter and body text. Full translation preserves all original arguments; digest mode mixes translation with commentary but must keep speakers clearly separated.

What does article_check.py validate in an article?▼

The script checks body length against prototype limits, banned template phrases, heading and bold counts for mobile readability, consecutive blank lines, paragraph rhythm, and first-person numeric claims. For translations it additionally verifies permission status and that source URL and author appear visibly in the body.

Can the Skill publish articles directly to WeChat or X?▼

No, it never auto-publishes. Finished drafts are handed off to ray-wechat or ray-x-article for platform draft creation, and the user confirms all publishing actions. Translations without granted permission stay local and never reach platform backends.

Does the writing workflow require an Obsidian knowledge base?▼

A compatible vault with .ray-obsidian.json or the standard directory structure enables the full pipeline with research archival and knowledge reflux. Without one, the Skill can deliver a one-off article in the current workspace, or hand off to ray-obsidian to initialize a vault first.

What happens when article_check.py reports warnings?▼

Warnings flag risks for human judgment, not automatic failures. Each warning is reviewed individually and can be reasonably retained if the article's natural flow justifies it; errors must be fixed and the check rerun before delivery.