feat(axe): add axe-core a11y gate (serious/critical, config-driven)#6
Merged
Conversation
gates/axe-gate.mjs runs axe-core over each built page in a real browser and
fails closed on any violation at or above a configurable impact threshold
(default: serious). Everything is config-driven (dist dir, page list, ruleset
tags, threshold, runner) with neutral defaults — nothing site-specific is
hardcoded. It serves dist over an ephemeral origin so absolute asset paths
resolve (running file:// would strip styles and fabricate layout-dependent
findings), then drives one of two interchangeable engines:
- playwright (default, CI): @axe-core/playwright + bundled Chromium
- tezcatl (local, macOS WebKit): injects axe.min.js and reads results back
The emitted machine-readable report groups violations by impact per page and
exposes an `axe: { serious, critical }` envelope — exactly what
conformance-report's `a11y.axe-serious-critical` criterion consumes, so a clean
run is what lets a site honestly assert it and a regression turns CI red.
Pure classification/threshold/report logic is exported and unit-tested
deterministically; a fixtures/axe/ known-bad + known-good pair is exercised
end-to-end when a browser engine is on PATH (skipped, like the cosign step,
when none is). axe-core added as a dependency; @axe-core/playwright + playwright
are consumer-supplied (dynamically imported) to keep `npm ci` light.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011uU1XvPggEPNMiXBUrV8hy
bdelanghe
added a commit
to bdelanghe/site
that referenced
this pull request
Jun 29, 2026
…it gate via tezcatl locally (#161) * refactor(conformance): consume the kit's axe gate (real browser) instead of the bespoke jsdom one Re-vendor conformance-kit to current main (dbc2c54) and replace bd-site's ad-hoc jsdom axe gate with the kit's reusable, dual-runner gate (bounded-systems/conformance-kit#6). Why it's a strict upgrade: the kit gate loads each built page in a REAL browser and runs axe's full WCAG 2.x A/AA ruleset, so it covers the layout-dependent rules (color-contrast, target-size) that jsdom forced us to disable. Verified clean over all 7 pages via tezcatl (macOS WebKit, no Chromium): 0 violations at any level — confirming the jsdom gate hid nothing, and axe now corroborates lone's static color_contrast check rather than splitting coverage. - scripts/vendor-integrity.mjs — add gates/axe-gate.mjs to the vendored FILES. - .github/workflows/conformance.yml — axe step now runs the vendored kit gate; CI uses the Playwright/Chromium runner (`npx playwright install`), local dev uses tezcatl (`npm run check:axe`, no Chromium download). - package.json — drop jsdom; add @axe-core/playwright + playwright; check:axe → kit gate via tezcatl. Remove scripts/axe-gate.mjs. - data/conformance-evidence.json — _gates/_source updated (real browser, full coverage). evidence.axe {serious:0, critical:0} unchanged. Side effect of the re-vendor: the kit's external-grader criteria (#5: scorecard, HSTS-preload, SLSA level) now appear on /conformance as not-assessed (bd-site supplies no evidence for them) — additive and honest. structure.json baseline regenerated for the new /conformance rows. SBOM self-heals (352 pkgs, check:sbom green). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * refactor: keep axe CI browser-free (jsdom), add the kit gate for local full-coverage via tezcatl Reworks the kit-gate adoption per the no-macOS / no-Chromium constraint. tezcatl is native macOS WebKit (links AppKit/CoreGraphics) — it cannot run on a Linux CI runner — so CI stays BROWSER-FREE on the bespoke jsdom gate, and the vendored kit gate is the DEEPER LOCAL pass (real browser via tezcatl, full WCAG 2.2 A/AA incl. contrast). - CI (conformance.yml): axe runs `scripts/axe-gate.mjs --check` (axe-core in jsdom, no Chromium, no macOS). color-contrast / target-size deferred to lone's static color_contrast check, as before. - Local: `npm run check:axe` runs the vendored kit gate via tezcatl — verified clean over all 7 pages incl. contrast. - Drop @axe-core/playwright + playwright (unused on this path); keep axe-core + jsdom. - _gates/_source document the two layers honestly. Net change vs main is just the re-vendor (kit→dbc2c54: brings the #5 external-grader criteria — now not-assessed on /conformance — and vendors the kit axe gate for local use) + pointing check:axe at it. structure.json regenerated for the new rows; SBOM self-heals (check:sbom green). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
bdelanghe
added a commit
that referenced
this pull request
Jun 29, 2026
… evidence) (#7) Promotes bd-site's bespoke vuln gate into the kit as a reusable, config-driven gate (the axe-gate #6 pattern): pure parse/evaluate core (unit-tested over synthetic npm-audit payloads) + a config-driven CLI + the ck-vuln-gate bin. Runs `npm audit` over a project's lockfile and FAILS CLOSED when the known critical/high count exceeds a threshold (default 0). Production-scoped by default ($VULN_OMIT_DEV, since a static site ships no runtime deps); $VULN_THRESHOLD, $VULN_ROOT, $VULN_REPORT configurable. The report's `vulns: {knownCriticalOrHighVulns}` envelope is exactly what lone's conformance() consumes for `security.no-critical-vulns`. Tests: pure parse/threshold logic + a best-effort real `npm audit` e2e (tolerated skip when offline). node test/run.mjs → 14 passed, 0 failed. NOTE: the e2e surfaced 1 high advisory in the kit's OWN production deps — a real supply-chain finding for the kit, tracked separately. Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This was referenced Jun 29, 2026
bdelanghe
added a commit
that referenced
this pull request
Jun 29, 2026
…mance evidence) (#10) Third and final gate-promotion (epic prx-em8h). Promotes bd-site's bespoke baseline gate into the kit, same pattern as vuln (#7), vnu (#9), axe (#6). Maps the SHIPPED CSS to web-features Baseline data via stylelint-plugin-use-baseline (headless, no browser) and FAILS CLOSED when the site-wide status is below target. HONEST: reports the MEASURED status (widely/newly/limited = the worst feature used); a feature guarded behind an @supports query is a tested fallback and doesn't count. Config-driven ($BASELINE_CSS/$BASELINE_TARGET/$BASELINE_REPORT); no site coupling. Emits the `baseline: {status, fallbackTested}` envelope lone's conformance() consumes for compatibility.baseline. Unlike vnu/axe, the engine (stylelint) is pure npm, so the fixture e2e is DETERMINISTIC and runs in CI (no skip). Tests: pure classify/threshold logic + e2e over good (widely) / bad (:has + ::selection → limited) fixtures. node test/run.mjs → 16/0. With this, all three bd-site gates are reusable in the kit; bd-site can now consume them vendored (prx-v5ry), and the kit is ready to publish (prx-47qm). Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds
gates/axe-gate.mjs— a generic, config-driven axe-core accessibility gate so the conformance model'sa11y.axe-serious-criticalcriterion becomes a continuously-ENFORCED, gate-backed check instead ofnot-assessed.What it does
serious).$DIST/ argv (dist dir),$AXE_PAGES(default: every*.html),$AXE_TAGS(defaultwcag2a,wcag2aa,wcag21a,wcag21aa,wcag22aa),$AXE_IMPACT_THRESHOLD,$AXE_RUNNER,$AXE_REPORT.file://would strip styles and fabricate layout-dependent violations).@axe-core/playwright+ bundled Chromium) and tezcatl (local, macOS WebKit — injectsaxe.min.js, reads results back).axe: { serious, critical }envelope thatconformance-report's evaluator consumes verbatim.Tests
fixtures/axe/{bad,good}.htmlexercised end-to-end when a browser engine is on PATH (skipped, like the cosign step, when none is). Local run: known-bad = 1 critical / 2 serious blocking, known-good = clean. 13/13 pass.Packaging
axe-coreadded as a dependency (small; used by the tezcatl injection path).@axe-core/playwright+playwrightare consumer-supplied (dynamically imported with a clear error) to keepnpm cilight — sites install them in theiraxe.yml.Independent PR; based on
main. Sites (bdelanghe/site,bounded-systems/site) re-vendor this gate + wire anaxe.ymlworkflow in companion PRs.🤖 Generated with Claude Code