Ray Wang avatar

Ray Wang

Community

@imraywang · on earch

60Followers
|
13Public Repos
|
10Published Skills

Founder @Upthos

Skills Distribution
DomainGrowth, Mark...Chinese Long-Form .. (30%)Multi-Platform Dra.. (25%)Spoken-Video Scrip.. (20%)Obsidian Knowledge.. (15%)

Agent Skills by Ray Wang

Showing 10 vetted skills indexed across 1 GitHub repositories.

Frequently Asked Questions About Ray Wang

FAQPage Schema
What tasks can I accomplish with Ray Wang's rayskills?▼

You can run an end-to-end Chinese content pipeline: build an Obsidian knowledge base, write or translate long-form articles, generate covers, produce spoken-video scripts and collage B-roll videos, and deliver drafts to WeChat official accounts and X Articles. Skills never auto-publish; platform writes require explicit confirmation.

Who are the rayskills designed for?▼

They target Chinese-language content creators, newsletter writers, and solo publishers who produce long-form articles for WeChat and X, repurpose them into short spoken videos, and manage source material in a local Obsidian vault. ray-consult additionally serves enterprises evaluating knowledge-base or RAG adoption readiness.

How do the skills handle publishing to WeChat and X?▼

ray-wechat and ray-x-article only create or update drafts after explicit user confirmation; they never click publish or mass-send. Ambiguous commands like 'push to platform' first produce a local preview package, then ask a one-line confirmation before writing to any draft box, avoiding duplicate same-title drafts.

What are the prerequisites for using the writing pipeline?▼

ray-writer requires a compatible local Obsidian knowledge base; if none exists, it hands off to ray-obsidian, which incrementally builds the vault structure without overwriting existing notes or forcing plugins, sync services, or Git. Platform delivery skills need a logged-in WeChat or X session.

How does ray-multimodel coordinate multiple models?▼

It assigns minimal-sufficient task splits across Grok, Claude, and Codex with the current session acting as verifying controller. Five entry modes exist: auto, fast, review, race, and scout, where scout uses an isolated Grok 4.5 search flow to gather X, Reddit, and web evidence. It is skipped for simple Q&A.