webtest-orch v0.3.2-beta

Documentation-only redaction release.

Changed

• Removed direct reference to lingua.creatman.site in the validation section of README.md and README.ru.md. Lingua Companion is in private beta — its production domain should not appear in any public webtest-orch artefact until product launch. The github repo link (github.com/CreatmanCEO/lingua-companion) remains; the domain link is replaced with a (private beta) descriptor inline.
• 0.3.1-beta is now deprecated on npm. Users should upgrade to 0.3.2-beta via npx webtest-orch@beta install.

Note

No functional changes. Same 113 tests. Same CI matrix.

webtest-orch v0.3.1-beta

Documentation-only re-positioning release. No functional changes; same 113 tests, same CI matrix.

Changed

• README rewritten around honest positioning: "token-efficient e2e orchestration skill for Claude Code." Hook leads with verified token-cost numbers (Playwright MCP ~1.5M tokens / Playwright CLI ~25–27k per Özal / TestDino / Morph benchmarks), real peer group is the free / OSS tier (playwright init-agents --loop=claude + Magnitude), explicit honest table separating where the skill genuinely competes vs where claiming "alternative to Octomind / QA Wolf" would overstate. Validation section now links to the actual public OSS apps it was tested on (CreatmanCEO/portfolio, CreatmanCEO/lingua-companion).
• README.ru.md rewritten with the same positioning frame.
• reference/playwright-patterns.md "Healing policy" section clarified — webtest-orch deliberately does NOT ship self-healing. Rationale: the QA community has pushed back on self-healing as marketing spin (false-green from visually-similar-but-wrong elements). Microsoft's native init-agents --loop=claude Healer is free, opt-in, and our generated specs are compatible — recommended skip-real-bugs policy documented.
• Repo description updated to: "Token-efficient e2e orchestration skill for Claude Code: explore once, replay deterministically. Playwright + axe-core + run-diff. Tests stay in your repo. MIT."
• Topics swapped: removed claude-code-config and image-budget, added test-orchestration and playwright-testing for discoverability via the dominant 2026 search terms.

Added (docs only)

• "Built on real benchmarks" section in README — every architectural choice cited to a verifiable source (Özal benchmark on GitHub, TestDino, Morph, Microsoft Fara-7B paper, arXiv 2511.19477, Deque WCAG study, W3C, Anthropic skill spec).
• "Related work and credible voices" section linking to Simon Willison's TIL, Matt Pocock's skills repo (45k★), Alexander Opalic's "AI QA Engineer" tutorial, Pramod Dutta's token-cost analysis, and Microsoft's init-agents docs.
• "What we deliberately do NOT ship" section — explicit on no self-healing, no vendor cloud, no "AI writes all your tests" pitch.

webtest-orch v0.3.0-beta

Distribution milestone — installable via npx webtest-orch@beta install instead of git-clone-and-bash.

Added

• npm package webtest-orch published to npm registry under beta tag.
  Install path: npx webtest-orch@beta install (no global install needed).
• bin/webtest-orch.js — CLI with subcommands:
  • install — copy skill into ~/.claude/skills/webtest-orch/
  • install --symlink — symlink for local development
  • uninstall — remove installed skill (npm package untouched)
  • status — show install state + MCP availability
  • version, help
• .github/workflows/release.yml — on v* tag: lint + tests, extract release
  notes from CHANGELOG, create GitHub Release (prerelease for beta/alpha/rc),
  publish to npm with provenance. Requires NPM_TOKEN repo secret to push to npm.

Changed

• Skill name renamed webapp-test-orchestrator → webtest-orch in SKILL.md
  frontmatter. Installs into ~/.claude/skills/webtest-orch/. Existing
  ~/.claude/skills/webapp-test-orchestrator/ directories from 0.1.x/0.2.x
  are left in place (delete manually if not needed).
• README install section rewritten around the npm one-liner. Older bash install.sh
  flow documented as the alternative for development / no-npm setups.