Scanning

Basic usageCopy Anchor Link

equall scan [path]

path defaults to the current directory. Exit codes:

A successful scan exits 0 by design, so an interactive run never looks like a failure. Opt into a CI gate with --min-score, or gate a pull request on what it introduces with --diff and --fail-on, which has its own exit codes.

An existing folder with no scannable files prints "No scannable files found" and exits 0; check the path if a CI step reports it. A path that does not exist exits 2 with path not found, so a typo in a CI step fails instead of passing.

Warnings about what the scan could not check (a file a scanner could not parse, a scanner that failed) are printed on stderr before the report.

Support SummaryCopy Anchor Link

The terminal report ends with its takeaway — printed last so it's what stays on screen when the scan finishes. The 0–100 score (a trend indicator, not a pass/fail grade) and an honest verified-subset statement sit just above a WCAG 2.2 Support Summary — three buckets, derived from criterion_conformance, that sum to the target level's criteria total:

  ◆ EQUALL — Accessibility Score

  Summary
  3 files scanned  ·  7 WCAG violations  ·  0 best-practice recommendations  ·  4 page-level (needs rendered page)
  ■ 3 critical   ▲ 4 serious   ● 0 moderate   ○ 0 minor
  critical/serious = fix before shipping · moderate/minor = fix in next iteration
  1 route detected · html 1
  1 finding not counted — static analysis can't confirm it · run with --show-review

  [...violations and details...]

    65.38    WCAG 2.2 · score is a trend indicator
  5 A/AA failures among the 20 criteria automatically verified (35 not evaluated).

  WCAG 2.2 Support Summary — AA target · automated basis only
  ✓ Supports (automated) 15   ✕ Does not support 5   ○ Not evaluated 35
  Automated verdicts only — a full statement needs manual + assistive-tech testing. Run --verbose for the per-criterion table.
  What each verdict means → https://equallscan.com/docs/verdicts

"Supports (automated)" and "Does not support" count criterion_conformance entries with verdict pass_automated and fail; "Not evaluated" is every other verdict combined. --verbose expands the full per-criterion table and splits "Not evaluated" into its three reasons — rendered check, assisted, or manual. The headline line above it is derived from the same verdicts: the criteria "automatically verified" are the fail plus pass_automated ones, so the two lines always agree. A scan with no failures states this verified subset, never "None". None of the words "Meets", "conformant", "compliant", or "conformance" appear in the output — automated verdicts state a basis, not a certification.

The Summary counts only the issues that affect the score. Suppressed issues (equall-ignore) and findings static analysis cannot confirm are announced on their own lines, never counted as violations.

Findings to reviewCopy Anchor Link

On component source (JSX/TSX, Vue, Svelte, Astro), axe-core sees markup reconstructed from the code, not a rendered page. When a finding sits on an element whose name or content depends on props, runtime expressions or a component (<Button>, {...props}, aria-checked={state}), or comes from a rule that needs the rendered page, static analysis cannot confirm it. Such a finding is reported for review: it keeps its fingerprint and stays in the --json output (review_only: true, with a review_reason), but it never counts in the score, the violation counts or the Support Summary. Plain .html is never affected.

By default the terminal prints one line with their count. --show-review lists them in a "To review" section:

  To review — 1 finding static analysis can't confirm on component source · not counted

  ○ button-name  1 occurrence  WCAG 4.1.2
      The element depends on props, runtime expressions or a component, which static analysis cannot see.
      ↳ src/components/Toolbar.tsx

FlagsCopy Anchor Link

Common recipesCopy Anchor Link

Target Level A only

equall scan ./public --level A

Export a JSON report

equall scan . --json > report.json

Scope to a subfolder

equall scan . --include "src/**"

Run in CI (fail a pull request only on what it introduces)

equall scan . --diff origin/main --fail-on critical

Existing violations never block; see CI for the snippets per CI provider.

Run in CI (fail the build below a score threshold)

# .github/workflows/a11y.yml
- run: npx equall-cli scan . --level AA --min-score 90

Without --min-score, the scan reports the score but still exits 0 — handy for an informational CI step that never blocks the build.

Save a plain-text report

equall scan . --all > report.txt

The output is not a terminal, so colors are off and the file carries no escape codes. --all lifts the default caps so every occurrence is in the file.

Skip readability checks for non-English projects

equall scan . --no-readability

Evaluate against the WCAG 2.1 legal bar instead of 2.2

equall scan . --standard wcag21

--standard is a view filter on the conformance table, coverage, and verdict — the 0–100 score is identical regardless of which standard is selected. Under wcag21, the 9 success criteria new in 2.2 (e.g. 2.5.8 Target Size) leave the conformance table (findings on them remain visible as issues elsewhere in the output), and 4.1.1 Parsing reappears as a documented automated pass (obsolete per W3C erratum, so nothing new to check).

Verifying page-level rulesCopy Anchor Link

Landmark, document-title and language rules can't be judged from a single component — scan your built output (dist/) to verify them on the composed page. See Verifying page-level rules.

Ignore subcommandCopy Anchor Link

Managing inline equall-ignore comments also ships with a dedicated subcommand — see Ignore system.