From scripts to subtitles: 55 agent skills for content automation

What it does rnskill is a catalog of 55 SKILL.md files that teach agentic coding tools like Codex and Claude Code how to perform specialized content-creation tasks. The skills span the full video-production chain: topic planning, script rewriting, ASR transcription, local voice cloning, digital-avatar generation, rough cutting, subtitle timing, motion graphics, and even post-mortem analytics. A subset of generic coding skills is included, though the repository is overwhelmingly aimed at creators running short-form video pipelines.

The interesting bit The project treats AI agents like a production crew rather than a chatbot. Each skill is a narrow, structured contract—ra-人话 scrubs AI flavor from Chinese text, rn-cover-skill generates editorial SVG covers without reference images by splitting layout and concept generation into two layers, and ra-video-production-director orchestrates downstream skills with state management and QC gates. It is essentially an attempt to operationalize “vibe coding” for media workflows.

Key highlights

• 55 skills covering scripting, dubbing, editing, subtitles, motion graphics, and cover design
• rn-cover-skill generates fresh 5:2 editorial covers each time, outputting self-contained SVG and PNG with warm-white #FAF9F5 canvases
• Includes adapted commercial-diagnosis skills (dbs-*) from the dbskill project for content-quality checks
• Some skills integrate with specific tools: IndexTTS2 for local voice cloning, HeyGen for digital avatars, yt-dlp/TikHub for video download, and HyperFrames for motion graphics
• The maintainer claims the kit helped them gain 2,000 Douyin followers and land a first commercial order with under ten hours of monthly effort

Caveats

• Most skills are tightly coupled to a Chinese short-form video ecosystem (Douyin, Xiaohongshu, WeChat public accounts) and specific third-party services like HeyGen and IndexTTS2
• 22 of the 55 skills are adapted from the external dbskill project, so the repository is partly a curated redistribution with pipeline-specific tweaks
• The “under ten hours” claim is an anecdote from the maintainer, not a reproducible benchmark

Verdict Worth browsing if you are trying to automate a Chinese-language content pipeline with agentic tools and don’t mind gluing together yt-dlp, HeyGen, and local TTS. Skip it if you are looking for general-purpose coding assistance or a turnkey product that works out of the box.