Skip to content

Releases: yanstd888/humanizer-lint

humanizer-lint v1.2.1

Choose a tag to compare

@yanstd888 yanstd888 released this 31 Oct 07:25
27dcb76

Added

  • Visible Tier 1 analysis controls for English, Indonesian, Spanish, Malay, French, German, Portuguese, and Mandarin, plus Universal Stylometry for 100+ languages.
  • Shareable targetLang playground URLs and six multilingual AI samples with deterministic failures.
  • Static multilingual coverage, use cases, FAQ markup, and structured data for search and agent discovery.

Changed

  • SKILL.md, README, npm metadata, LLM discovery files, sitemap, and social preview now describe the full multilingual capability.
  • Auto-fix prompts identify the active language and rule pack.

Verification

  • Node.js 18, 20, and 22 CI passed.
  • Cloudflare Pages production was verified on the custom domain and Pages mirror.

humanizer-lint v1.2.0

Choose a tag to compare

@yanstd888 yanstd888 released this 31 Oct 06:16
5e23aa0

Added

  • Universal multi-language sentence and word segmentation through native Intl.Segmenter.
  • Built-in anti-slop packs for English, Indonesian, Spanish, Malay, French, German, Portuguese, and Mandarin.
  • Universal stylometric fallback for other ISO 639-1 inputs, including CJK and Arabic script detection.
  • Analysis-language selector in the local browser playground.

Changed

  • CLI accepts --lang auto, --lang universal, or a two-letter ISO 639-1 code and reports the active rulePack.
  • Static discovery documents, agent metadata, package whitelist, and README now describe the v1.2.0 language tiers.

Verification

  • Node.js 18, 20, and 22 CI passed.
  • Production Pages deployment verified on the official domain and Pages mirror.

v1.1.0

Choose a tag to compare

@yanstd888 yanstd888 released this 31 Oct 04:06
8755abe

Added

  • Fixture-backed interactive browser playground samples, metric cards, and clickable diagnostics.
  • Static SEO/GEO assets: JSON-LD, Open Graph image, robots, sitemap, and LLM discovery files.
  • Agent-discovery metadata, deep-link documentation, and Node 18/20/22 CI coverage.

Changed

  • Browser playground remains local-first and zero-dependency.

Compatibility

  • Analyzer engine, scoring, rules, and profiles are unchanged.

humanizer-lint v0.2.4

Choose a tag to compare

@yanstd888 yanstd888 released this 30 Oct 20:34

Added

  • Published CLI installation and npm documentation.
  • Static browser demo at humanizer-lint.pages.dev.

Changed

  • Shared analyzer and metrics core between CLI and browser demo.
  • Added CLI↔browser parity coverage and deterministic demo build.
  • Added Indonesian before/after documentation example.

Verification

  • Node.js 18.3.0 and Node.js 22 CI checks pass.
  • npm test: 19/19 pass.
  • Production demo verified in desktop and mobile browser checks.

v0.2.3

Choose a tag to compare

@yanstd888 yanstd888 released this 30 Oct 17:26

Fixed

  • Replaced the temporary static skills.sh badge with the official install-count badge format documented by skills.sh.
  • Pointed the badge to the repository-level skills.sh page, removing the repeated skill slug from the README link.
  • Added a regression check for the exact official badge markup and repository-level URL.

Verification

  • 18 of 18 tests passed.
  • Official Agent Skills validation passed.
  • Package dry-run produced 12 intended public files.
  • GitHub Actions passed on Node.js 18.3.0 and Node.js 22.

v0.2.2

Choose a tag to compare

@yanstd888 yanstd888 released this 30 Oct 17:05

Fixed

  • Replaced the broken skills.sh dynamic badge that rendered as custom badge: resource not found.
  • Added a working skills.sh: listed badge linked directly to the published humanizer-lint skill page.
  • Added a regression check that rejects the broken badge endpoint.

Verification

  • 18 of 18 tests passed.
  • Official Agent Skills validation passed.
  • Package dry-run produced 12 intended public files.
  • GitHub Actions passed on Node.js 18.3.0 and Node.js 22.

v0.2.1

Choose a tag to compare

@yanstd888 yanstd888 released this 30 Oct 16:56

Changed

  • Rebuilt the README as an English project landing page with source-verified commands, metrics, scoring rules, limitations, and deterministic terminal examples.
  • Added the verified skills.sh installation path, badge, and an accurate explanation of GitHub's native Used by panel.
  • Documented the current zero-dependency CLI and voice fingerprint workflow without claiming unsupported npm publishing, auto-fix, directory scanning, SARIF, or custom config files.

Added

  • Added a minimal GitHub Actions workflow that runs tests and package verification on Node.js 18.3.0 and Node.js 22.
  • Added regression checks for the README contract, skills.sh badge, no-emoji policy, workflow commands, and synchronized version metadata.

Verification

  • 18 of 18 tests passed.
  • Official Agent Skills validation passed.
  • Package dry-run produced 12 intended public files.
  • Public CLI smoke tests returned PASS score 100 with exit 0 and FAIL score 20 with exit 1.
  • GitHub Actions passed on Node.js 18.3.0 and Node.js 22.

v0.2.0

Choose a tag to compare

@yanstd888 yanstd888 released this 30 Oct 16:32

Added

  • Public humanizer-lint and humanizer-fingerprint CLI names.
  • Indonesian/English function-word language detection for technical prose.
  • Custom voice-profile deviation diagnostics and scoring.
  • Human-readable terminal output alongside deterministic JSON output.

Fixed

  • Markdown links now preserve anchor text without counting URL syntax.
  • Indonesian reduplication remains one token; decimals, thousands, abbreviations, and ellipses no longer split sentences incorrectly.
  • Indonesian loanwords ending in common borrowed suffixes no longer trigger passive-voice false positives.
  • Translation-style subordinate openings are checked only when consecutive inside the same paragraph.
  • --stdin fails fast when no piped input is present.

Security and packaging

  • Removed the internal build blueprint from the public repository.
  • Added an MIT license, official Agent Skills frontmatter, zero-dependency package allowlist, and static checks blocking network, shell execution, eval, and dynamic function construction.
  • Verified with 18 Node tests, the official agentskills validator, CLI smoke tests, and npm pack --dry-run.