skills

Reusable Codex skills for browser automation, page extraction, screenshots, note workflows, audio/video pipelines, publishing flows, and Windows local tooling.

Install

Clone once:

git clone --depth 1 https://github.com/picasso250/skills.git ~/tmp/picasso-skills

Copy one skill into your local skills directory:

cp -r ~/tmp/picasso-skills/website-to-cli ~/.agents/skills/

Replace ~/.agents/skills/ with your actual skills directory if different. The examples above use bash-style paths and commands; adjust them if you are installing from PowerShell or another shell.

After copying, check whether the installed skill directory contains a README.md. If it does, follow that README to finish installation.

Update

Refresh the local clone:

cd ~/tmp/picasso-skills
git pull

Then replace the installed skill with the updated one.

Skills

Current installable skills in this repo:

• ai-speak: real-time AI voice playback and self-talk mode; not for exporting wav, srt, or video assets.
• audio-to-srt: generate acoustically aligned srt/json subtitle timelines from wav/mp3 audio.
• bilibili-downloader: extract audio/video streams from an already-open Bilibili page and optionally save local m4a or muxed mp4.
• bilibili-video-maker: generate a Bilibili-ready video pipeline from a topic, including script, AI voice, and animated subtitles.
• continue-on-phone: continue the current agent conversation on a phone by using the local QR flow.
• desktop-screenshot: capture the current Windows desktop to a PNG file.
• douyin-manage: extract first-screen work list data from the Douyin creator content management page.
• gemini-img-gen: generate high-resolution images with Gemini image generation.
• handover: compress the current task state into exactly five sentences and write handover.md.
• materialist-dialectics: run a fact-grounded thesis-antithesis-synthesis conversation when the user explicitly asks for that mode.
• notes-workflow: write notes into the local Markdown notes repository under $HOME\Documents\notes.
• ollama-ocr: run local OCR with Ollama multimodal models to extract text, tables, or document structure.
• screenshot-webpage: capture screenshots of web pages or local HTML files, preferably through an existing remote-debug browser session.
• send-email: send email to the user.
• shared-page-md: reuse the current browser session and convert a page DOM into Markdown.
• text-to-wavs: generate wav audio plus srt/json subtitle timelines from text.
• v2ex: read a V2EX topic and export a Markdown snapshot with metadata, body, and replies.
• website-to-cli: turn a repeatable browser workflow into a deterministic CLI flow by pairing Chrome DevTools with Playwright.
• wechat-mp-publish: automate WeChat Official Accounts publishing through a staged browser workflow.
• windows-uac: handle Windows tasks that require elevation, preferring direct Start-Process -Verb RunAs when admin rights are clearly needed.

Notes

• The list above intentionally tracks installable skill directories that contain a SKILL.md.
• website/ contains site assets, not an installable skill.
• browser/ is helper/runtime material rather than a standalone installable skill directory.
• message-search/ currently has no installable skill manifest and is not listed above.
• Some skills include their own README.md or extra setup notes. After copying a skill, read that skill's local documentation before use.