Lighthouse Scanner

lighthouse-scanner Skill

Integration patterns for Lighthouse CI accessibility auditing (treosh/lighthouse-ci-action). Teaches agents how to detect Lighthouse CI configuration, parse accessibility reports, map Lighthouse weight-based severity to the standard model, correlate with local axe-core scans, track score regressions, and produce structured JSON output.

Agents That Use This Skill

Agent	Why
lighthouse-bridge	Primary consumer -- bridges Lighthouse CI data into the agent ecosystem
web-accessibility-wizard	Auto-detects Lighthouse in Phase 0, correlates findings in Phase 9
insiders-a11y-tracker	Discovers Lighthouse-related issues and score regressions
daily-briefing	Surfaces Lighthouse score regressions in daily CI Scanner report section
issue-tracker	Recognizes Lighthouse CI issues in Scanner Triage mode

What the Skill Covers

Lighthouse CI Detection

How to check if a repository uses Lighthouse CI:

• Grep .github/workflows/ for treosh/lighthouse-ci-action or lhci autorun
• Check for config files: lighthouserc.js, lighthouserc.json, .lighthouserc.js, .lighthouserc.json, .lighthouserc.yml
• Parse config for URLs, numberOfRuns, score thresholds, and upload targets

Accessibility Score Interpretation

Lighthouse produces a 0-100 accessibility score per URL. The skill maps scores to grades:

Score Range	Grade
90-100	A
80-89	B
70-79	C
60-69	D
0-59	F

Severity Mapping

Maps Lighthouse audit weights to the standard agent severity model:

Lighthouse Weight	Agent Severity	Weight
10	Critical	10
7	Serious	5
3	Moderate	2
1	Minor	1

axe-core Rule Correlation

Lighthouse accessibility audits are powered by a subset of axe-core. The skill maps Lighthouse audit IDs directly to axe-core rule IDs for deduplication with local scans. Key mappings include:

• image-alt, color-contrast, label, button-name, link-name
• html-has-lang, document-title, meta-viewport, heading-order, tabindex

Score Regression Detection

Tracks accessibility scores across CI runs and classifies changes:

• Score drops of 10+ points: Critical regression
• Score drops of 5-9 points: Serious regression
• Score drops of 1-4 points: Moderate regression
• Score improvements are also tracked for positive trend reporting

Confidence Boosting

Findings confirmed by multiple sources get confidence upgrades:

• Lighthouse + local axe-core = High confidence
• Lighthouse + agent review = High confidence
• Lighthouse + axe-core + agent review = Highest confidence
• Lighthouse only = Medium confidence

Full Skill Reference

See the complete skill file: .github/skills/lighthouse-scanner/SKILL.md