thesmythgroup/Smyth Accessibility
LLM-based first-pass accessibility analysis on web-related code (PR changeset or full repo)
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| scope | Which set of files to analyze: pr (changeset only) or full (entire repo) | yes | pr |
| provider | LLM provider: openai or anthropic | yes | openai |
| model | Model name (e.g. gpt-4o, claude-3-5-sonnet) | yes | gpt-4o |
| acceptability-level | Acceptability threshold: wcag-a, wcag-aa, wcag-aaa, or custom | no | wcag-aa |
| file-patterns | Comma-separated globs or extensions (default: web-related). Example: "*.tsx,*.jsx,*.html" | no | "" |
| base-ref | For PR mode only: override base ref for diff (default: PR base sha) | no | "" |
| fail-on | When to fail the job: none, warn, or error | no | warn |
| post-pr-comment | When true and event is pull_request, post a comment with the summary | no | false |
| github-token | Token for posting PR comments. Defaults to GITHUB_TOKEN from the runner; pass secrets.GITHUB_TOKEN from your workflow if the warning appears. | no | "" |
Outputs
| name | description |
|---|---|
| findings-count | Total number of findings (errors + warnings) |
| has-errors | Whether any error-level findings were reported |
| has-warnings | Whether any warning-level findings were reported |