v1.4.1 - Post-fix ruthlessness gates

Adds mandatory post-fix ruthlessness checks to the Codex wrapper, installer prompt, default prompt, Spark workflow bundle, proof packet contract, and key H70 design lenses. The guild now explicitly checks after screenshots for awkward leftover onboarding, duplicated actions, floating copy, box soup, text-first clutter, and responsive regressions before accepting a visual pass.\n\nValidation run:\n- npm test\n- h70-cplus-convention-auditor on visual-loop-qa, user-flow-friction-auditor, responsive-vision-auditor, component-system-steward

v1.4.0

Adds the product-intent-observer lens and makes product/page intent the first-class starting point for important visual work. Also tightens the Codex wrapper around viewport coverage, vision-backed handoffs, flow clarity, and box-soup visual checks. Validation: npm test, H70-C+ convention auditor, npm run install:codex.

v1.3.0 - Flow and component-system guild lenses

Adds two workflow hardening lenses: user-flow-friction-auditor for first-time clarity, overwhelm, next-action, and recovery checks; and component-system-steward for art-bible, token, and reusable component-system compliance. Updates the Codex wrapper, bundle load order, prompts, quickstart, workflow recommender, proof packet scaffolds, and tests to route these lenses without adding ceremony. Validation: npm test passes with 18/18 H70-C+ skills.

v1.2.2 - Visual workflow regression guards

What changed

• Removed stale desktop/mobile-only wording from workflow docs, benchmark template, no-ceremony prompt, and proof packet generator.
• Updated helper outputs to require desktop, tablet, mobile, and awkward fluid-width screenshots with Codex App vision observations.
• Hardened the usage audit with negative assertions that fail if stale desktop/mobile-only or generic vision language returns.
• Tightened workflow recommender evidence for tablet/fluid responsive checks.
• Removed broad imagegen tags from non-asset specialist YAML skills so routing stays less noisy.

Validation

• npm test
• H70-C+ validator: 16/16 valid, 0 errors, 0 warnings
• smoke test
• usage audit
• local npm run install:codex

v1.2.1 - Codex App native visual workflow

What changed

• Made Codex App native capabilities explicit in the wrapper skill: in-app browser context, local screenshot paths, Codex App vision, and imagegen only after assets are integrated.
• Extended first-run and specialist prompts to require desktop, tablet, mobile, and awkward fluid-breakpoint screenshots.
• Added vision-backed handoff packets and native guardrails to README, PROMPTS, QUICKSTART, installer output, and the demo proof packet.
• Tightened usage audit coverage so installer, docs, metadata, proof packet, and demo cannot drift back to desktop/mobile-only or non-vision claims.

Validation

• npm test
• H70-C+ validator: 16/16 valid, 0 errors, 0 warnings
• smoke test
• usage audit
• local npm run install:codex

v1.2.0 - Codex vision proof scoring

What changed

• Made actual Codex App vision observations part of the required Run Report Contract.
• Added npm run score-proof-packet for a 0-100 evidence score across screenshots, vision observations, viewport coverage, issue selection, fix specificity, verification, residual risk, and automation notes.
• Updated proof packet creation/checking, installer prompt, Codex wrapper, OpenAI metadata, README, QUICKSTART, PROMPTS, WORKFLOW, FREE-DROP, and demo proof packet.
• Added benchmarks/ with a no-fake-results guild-vs-plain-Codex benchmark protocol and run template.
• Kept Playwright as an optional scaffold only; no Playwright dependency was added to the default install.

Validation

• npm test
• H70-C+ validator: 16/16 valid, 0 errors, 0 warnings
• smoke test
• usage audit, including workflow recommender, proof checker, proof scorer, and Playwright scaffold checks
• local npm run install:codex

v1.1.7 - Executable visual proof workflow

What changed

• Added npm run recommend-workflow to route different design workflow needs into evidence lanes, lenses, commands, and proof checks.
• Added npm run create-proof-packet to scaffold the Run Report Contract in a target app.
• Added npm run check-proof-packet to fail blank or evidence-free proof packets and verify screenshot file paths.
• Added npm run scaffold:playwright-visual to create a starter Playwright toHaveScreenshot baseline and mask stylesheet.
• Updated README, QUICKSTART, PROMPTS, WORKFLOW, FREE-DROP, MANIFEST, and usage audit coverage.

Validation

• npm test
• H70-C+ validator: 16/16 valid, 0 errors, 0 warnings
• smoke test
• usage audit, including workflow recommender, proof packet, proof checker, and Playwright visual scaffold

v1.1.6 - Evidence-first workflow hardening

What changed

• Added WORKFLOW.md with a no-ceremony evidence-first workflow, lens router, stronger proof packet, and optional automation recipes.
• Updated Codex wrapper, prompt spellbook, quickstart, README, installer prompt, and demo proof packet to require viewport/state matrix, accepted visual change, and automation notes.
• Hardened usage audit coverage so the minimum useful pass and anti-ceremony lens routing stay protected.

Validation

• npm test
• H70-C+ validator: 16/16 valid, 0 errors, 0 warnings
• smoke test
• usage audit

v1.1.5

What's changed

This release hardens the guild around evidence-based design work, not aesthetic ceremony.

• Added a required Run Report Contract to the native Codex wrapper.
• Updated QUICKSTART.md with the exact report fields Codex must return.
• Updated PROMPTS.md, the Codex/OpenAI default prompt, and the installer prompt so first runs end with the report contract.
• Updated the first-run demo proof packet to show the report shape.
• Extended usage-audit coverage so the report contract cannot quietly disappear.

Run Report Contract fields

• goal
• screenshots inspected
• top issues
• chosen issue
• lens used
• exact fix
• verification
• still weak
• reusable rule

Validation

• npm test
• H70-C+ convention auditor: 16 PASS, 0 warnings, 0 errors

v1.1.4

What's changed

This release turns the dogfood/evaluation feedback into a sharper first-run path for builders coming from X.

• Added QUICKSTART.md with the 30-second, 2-minute, 5-minute path.
• Added a README trust strip and a top-level 30/2/5 explanation before the deeper guild lore.
• Added a shortest useful prompt to PROMPTS.md.
• Tightened the native Codex wrapper with a Minimum Useful Pass and Specialist Output Contract so lenses return evidence instead of ceremony.
• Updated the OpenAI/Codex default prompt and installer prompt to bias toward one fast visible win.
• Added examples/first-run-demo/PROOF_PACKET.md so users can see what a useful before/after result should look like.
• Added FREE-DROP.md public-drop promise language.
• Extended usage-audit coverage for quickstart, proof packet, public-drop copy, installer prompt, and wrapper output contract.

Validation

• npm test
• H70-C+ convention auditor: 16 PASS, 0 warnings, 0 errors