Back to search
sariclean
MarketplaceWorks with Claude CodeGitHub
Turn any social post into one clean, cited Markdown file. Agent Skill for Claude Code and any SKILL.md agent.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Checked today
Badge for your README
Markdown
[](/listings/alfindigital--sariclean)HTML
<a href="/listings/alfindigital--sariclean"><img src="/api/listings/alfindigital--sariclean/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
Claude Code:
/plugin marketplace add alfindigital/sariclean /plugin install sariclean@sariclean
Other agents (Codex, Cursor, Gemini CLI and others that read `SKILL.md`):
npx skills add alfindigital/sariclean
Or copy `skills/sariclean/` into your agent's skills folder by hand.
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
.claude-plugin/marketplace.json · Manifest · 16 lines
- {
- "name": "sariclean",
- "owner": { "name": "alfindigital", "url": "https://github.com/alfindigital" },
- "metadata": {
- "description": "SariClean: save social posts as cited Markdown notes."
- },
- "plugins": [
- {
- "name": "sariclean",
- "source": "./",
- "description": "Save any social post or article as one clean Markdown note: cited summary, original text, OCR, video transcript and top comments.",
- "version": "1.0.0",
- "license": "MIT"
- }
- ]
- }
README.md · README · 139 lines
- # SariClean
- Turn any social post into one clean, cited Markdown file.
- SariClean is an open-source [Agent Skill](https://agentskills.io) for Claude Code and any
- agent that reads `SKILL.md` folders. Give it a link from X, Threads, Reddit, YouTube,
- TikTok, LinkedIn, Instagram, Facebook, Telegram or any web page. You get one Markdown
- note with:
- - the original post, word for word, in its original language
- - OCR of its images and a timestamped transcript of its video
- - up to 30 useful top comments, with spam and usernames removed
- - a short summary where **every point cites its source**
- *Sari* is Indonesian for "essence".
- 
- ## Why
- Bookmarks do not help. You forget them, and your AI agent cannot read them. A folder
- of plain Markdown notes works with any editor, any search tool and any agent.
- ## Two rules
- 1. **No source, no point.** Each summary point carries a tag: `[P]` post, `[C3]`
- comment 3, `[L1]` a linked page, `[M1]` an image, `[T01:23]` a video time. A check
- script rejects a note with an untagged point or a tag that points nowhere.
- 2. **No file is better than a made-up file.** If the agent cannot read the post (for
- example, a login wall), it writes nothing and tells you which routes it tried.
- It never logs in, likes, follows or buys access for you.
- ## Install
- Claude Code:
- ```
- /plugin marketplace add alfindigital/sariclean
- /plugin install sariclean@sariclean
- ```
- Other agents (Codex, Cursor, Gemini CLI and others that read `SKILL.md`):
- ```bash
- npx skills add alfindigital/sariclean
- ```
- Or copy `skills/sariclean/` into your agent's skills folder by hand.
- ## Use
- ```
- sariclean https://x.com/someone/status/1234567890
- sariclean max=50 https://www.reddit.com/r/somesub/comments/abc123/
- sariclean refresh https://youtu.be/abc
- ```
- Notes are saved under `~/Documents/SariClean/<topic>/`. Set `SARICLEAN_ROOT` to use
- another folder. A manifest in `<root>/.sariclean/` remembers what you saved, so the
- same link is not saved twice.
- ## Example note
- ```markdown
- ---
- sariclean: 1
- url: https://x.com/someone/status/1234567890
- platform: x
- author: "@someone"
- posted: 2026-11-05
- captured: 2026-11-17T10:00+07:00
- status: ok
- read_with: [web-fetch, browser]
- comments_saved: 3
- gaps: []
- ---
- # Two prompts that make a site readable by AI
- ## Summary
- - The author suggests adding two prompts to the site repo [P]
- - A commenter asks for proof; the author points to the dashboard [C1] [C1.1]
- - The full prompts are on the linked page [L1]
- ### Try it
- 1. Run the first prompt on a public site repo and ask for a plan first [P]
- ### AI notes
- - The traffic number in the image is not verified.
- ## Post
- all you need are these 2 prompts, just add them to your codebase
- [M1] image: https://example.com/media/dashboard.jpg
- > OCR (machine, unchecked): AI responses 28.1K
- ## Comments
- ### 1
- Now prove it instead of posting a random image.
- ### 1.1 [author]
- check the dashboard
- ### 2
- Which model did you test this with?
- ## Links
- - [L1] https://example.com/prompts/ai-readable-site.md - the two prompts in full
- ```
- ## What it needs
- - **Required:** Python 3.9 or newer. The helper script uses the standard library only.
- - **For reading pages:** any web fetch tool your agent has. Comments on X, Threads,
- Instagram, Facebook and LinkedIn usually need a browser you are signed in to that
- the agent can drive, such as Claude in Chrome.
- - **Optional, for media:** `ffmpeg` and
- `pip install yt-dlp faster-whisper rapidocr-onnxruntime`. Without them, SariClean
- still saves the text and lists the missing media under `gaps`.
- See [skills/sariclean/references/sources.md](skills/sariclean/references/sources.md)
- for what works on each platform.
- ## Helper script
- ```bash
- python skills/sariclean/scripts/sariclean.py decide <url> # collect, skip, resume, retry, repair
- python skills/sariclean/scripts/sariclean.py check <note.md> # validate a note
- python skills/sariclean/scripts/sariclean.py finalize <note.md> # validate and add to the manifest
- python -m unittest discover -s skills/sariclean/scripts # run the tests
- ```
- ## Fair use
- SariClean is for your own reading archive. It reads only what you could see yourself,
- removes commenters' usernames, and does not keep downloaded media. Respect each
- platform's terms, and do not use it for bulk scraping or to republish other people's
- posts.
- ## License
- MIT. This is an independent project. It is not made by or affiliated with Anthropic.
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.