Back to search
raullenchai-skills
MarketplaceWorks with Claude CodeGitHub
Claude Code skills hand-crafted by Raullen from real day-to-day problems. More added over time.
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/raullenchai--skills)HTML
<a href="/listings/raullenchai--skills"><img src="/api/listings/raullenchai--skills/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.
Add the collection once:
/plugin marketplace add raullenchai/skills
Then install the skills you want:
/plugin install adjacent-possible@raullenchai-skills
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 · 12 lines
- {
- "name": "raullenchai-skills",
- "owner": { "name": "raullenchai" },
- "description": "Claude Code skills hand-crafted by Raullen from real day-to-day problems. More added over time.",
- "plugins": [
- {
- "name": "adjacent-possible",
- "source": "./plugins/adjacent-possible",
- "description": "Break out of your X (Twitter) filter bubble: prune noise, keep relationships, follow verified accounts one step outside what you already see."
- }
- ]
- }
README.md · README · 96 lines
- # skills
- A personal collection of [Claude Code](https://claude.com/claude-code) skills, hand-crafted by
- **Raullen** ([@Raullen](https://x.com/Raullen)) out of real problems in his own day-to-day work.
- Each skill is something he built for himself first, used for real, and then cleaned up to share.
- More skills will be added over time — each one is a separate plugin you can install on its own.
- ## Install
- Add the collection once:
- ```
- /plugin marketplace add raullenchai/skills
- ```
- Then install the skills you want:
- ```
- /plugin install adjacent-possible@raullenchai-skills
- ```
- ## Skills
- | Skill | What it does | Status |
- |---|---|---|
- | [adjacent-possible](#adjacent-possible) | Break out of your **X (Twitter) filter bubble**: prune noise, keep relationships, follow verified accounts one step outside what you already see | v0.1 |
- | *more coming* | | |
- ---
- ## adjacent-possible
- **A skill for breaking out of your filter bubble on X (Twitter).**
- > *The adjacent possible* — Stuart Kauffman's term for everything one step away from where you are:
- > reachable, but not yet explored.
- Recommendation feeds keep showing you more of what you already like. After a while that is all you
- see. This skill moves your X graph **one step outward** — not to random topics, but toward what *you*
- say you want to understand — and it does it from your own browser, at a human pace.
- ### What happens
- ```
- 1 Collect Claude reads your recent posts and who you follow (your logged-in Chrome, read-only)
- → a two-line diagnosis of your bubble
- 2 Ask "If your feed could teach you one or two new things this year, what would they be?"
- 3 Prune Followed accounts sorted into noise / sources / relationships.
- Relationships (your company, partners, investors, friends) are kept by default.
- 4 Expand 4–8 accounts per thing you want to learn, each verified as real and still active,
- each with one line on why it connects to you — plus a few long-form entry points
- 5 Apply Dry run first, then follow/unfollow in small batches at human speed
- 6 Check (optional) two weeks later: did the feed actually change? expand one more ring
- ```
- You spend about five minutes choosing. Claude does the reading, sorting, verifying and clicking.
- ### Requirements
- - Claude Code with [Claude in Chrome](https://claude.ai/chrome) connected
- - Chrome logged in to the X account you want to rebalance
- Then say something like *"my X feed is an echo chamber, help me break out"* or run `/adjacent-possible`.
- ### Safety
- - Runs entirely in your own browser tab. No passwords, no cookies, no private APIs, nothing sent
- anywhere by the script.
- - Nothing is followed or unfollowed until you approve the exact list and see a dry run.
- - Human pace: at most 20 actions per run, 15–35 s apart, and it stops if X shows a limit warning.
- Keep it under ~100 actions a day.
- - A keep-list protects accounts you never want touched.
- Automating clicks on X is your own call under X's terms of service; this tool is built to behave
- like a careful person, not a bot farm.
- ### Without Claude
- [`xedge.js`](plugins/adjacent-possible/skills/adjacent-possible/scripts/xedge.js) also works by hand:
- open X, paste it into DevTools → Console, then
- ```js
- // on https://x.com/<you>/following
- xEdge.start('following'); /* wait */ xEdge.status(); copy(xEdge.result(0, 9999))
- // follow / unfollow, dry run first
- xEdge.start('apply', { follow: ['someone'], unfollow: ['noise1'], keep: ['friend'], dryRun: true })
- xEdge.result(0, 50)
- ```
- The judgement parts — what to cut, what to add — are what the skill brings.
- ## License
- MIT
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.