v0.2.3 — works out of the box (URL audit from npm, no double-runs)

v0.2.3 — the "works out of the box" release

Three fixes that make SeoFlow actually work for strangers installing from npm:

Fixed

• URL audit now works from the npm-installed CLI. The published package has no python/ dir, so npx @imsankz/seoflow audit <url> silently fell back to HEAD-only signals — empty title/H1/schema/link analysis for every user, always. Added a pure-Node fetch fallback (Node 18+ fetch + built-in parser). Zero Python deps, real signals. Verified from the packed tarball: Title: Example Domain, H1s: 1, schema/link checks all live.
• Double pipeline execution via the published CLI. esbuild bundles run.ts into dist/bin/cli.js, so the import.meta.url auto-run guard always matched inside the bundle and ran the whole pipeline twice — duplicate reports, duplicate brain writes, double AI calls. Auto-run now keys on the entry filename only. Verified: 1 run, 1 report.
• Clean Gemini CLI auth failures. Token-level auth errors (stale/ineligible credentials) now print one clear hint per run instead of dumping raw stderr mid-run; the fallback chain takes over silently.

Try it

npx @imsankz/seoflow audit https://yoursite.com

Zero config, zero API keys. The canonical short name npx seoflow goes live with the next npm publish (dual-publish script included — npm run publish:both).