← ClaudeAtlas

trust-chainlisted

Find unknown-but-credible accounts to follow on X by mining the amplification graph of people you already trust, then verifying candidates on track record. Use when you want to discover high-signal sources in a topic, or build a self-correcting list of who to trust for information. Discovery is automated; verification is a documented method you apply yourself.
feinanbang/x-trust-chain-skill · ★ 0 · AI & Automation · score 70
Install: claude install-skill feinanbang/x-trust-chain-skill
# trust-chain — find who to actually trust on X Most people new to X follow the obvious institutional accounts and miss the people doing the real work. This skill fixes that: start from **one or two accounts you already trust** in a topic, and find the credible people *they* amplify — then verify those candidates on whether their past calls came true, not on follower count. Two halves, and they are honest about which is which: 1. **Discovery — automated.** `scripts/discover.js` mines who your trusted "seed" accounts retweet / quote / reply to / mention, and ranks the resulting candidates. Accounts amplified by *more than one* of your seeds (cross-validated) rank highest. 2. **Verification — a method you apply.** Discovery gives leads, not verdicts. Whether a candidate has *earned* trust is judgment-driven and differs per person, so it is documented as a framework, not hard-coded. See `docs/verification-method.md`. ## Setup (one time) 1. `npm install` 2. Copy `.env.example` → `.env`, add your own X API keys (free tier is fine). Your keys never leave your machine — `.env` is gitignored. 3. Copy `config.example.json` → `config.json`, set your `seeds` (handles you trust) and `domain`. ## Run discovery ``` node scripts/discover.js ``` Outputs a ranked candidate list to `./output/`. Cross-validated candidates first. Re-runs reuse the cache (0 API reads) unless you pass `--force`. ## Then verify before you follow Take the top candidates and run them throu