network-optimizer-plus

Built on affaan-m/ECC by @affaan-m (249,178 stars, MIT). All credit for the original idea to them. This fork improves and repackages it; upstream license preserved in UPSTREAM_LICENSE.

A Claude Code skill that helps you clean up, grow, and refocus your X and LinkedIn network without making unsafe changes.

Built for people who want a useful network, warm ties, and a clear review list.

💡 Why

Large networks get hard to manage.

Old follows can hide the people who matter now. Weak data can also lead to bad choices. A job change, quiet profile, or missing follow-back does not mean a person has no value.

This skill scores each person from 0 to 100. It shows the facts behind each score. It drafts a plan, then waits for your approval.

⚡ Install

mkdir -p ~/.claude/skills/connections-optimizer && curl -fsSL https://raw.githubusercontent.com/network-optimizer-plus/network-optimizer-plus/main/skill/SKILL.md -o ~/.claude/skills/connections-optimizer/SKILL.md

That is all. The skill has one file and no extra packages.

🧭 Usage

Ask Claude Code:

Use connections-optimizer to review my X follows.

My goals are to meet more open source builders and keep in touch with past coworkers.
Use default mode.
Do not change anyone from my protected list.
Show me the full plan before any action.

Expected output:

Network Review

Protect
- Sam Lee | Score: 91 | Past coworker and recent replies

Keep
- Ana Cruz | Score: 84 | Fits your open source goal

Review
- Dev Tools Daily | Score: 55 | Useful topic, but activity data is old

Prune Candidate
- Closed account | Score: 12 | Account is gone

Add Candidate
- Priya Shah | Score: 78 | Open source builder with a real warm path

No action was taken.
Waiting for your approval.

The skill can also draft X DMs, LinkedIn notes, and email text. It never sends them on its own.

🔧 What we changed vs upstream

• Rewritten from Japanese into clearer, simpler English with a shorter, more direct description.
• Adds stricter consent safeguards: no follow, unfollow, removal, blocking, or outreach without explicit approval.
• Adds uncertainty and data-integrity rules, including not inventing private data and skipping unclear identities.
• Refines platform guidance to avoid weak assumptions, such as treating no follow-back, job changes, or inactivity as proof of low value.
• Makes scoring more concrete by introducing a 0–100 score while preserving the original positive-signal framework.

📜 License

Released under the MIT License.

The upstream MIT license and credit are preserved in UPSTREAM_LICENSE.