| name | seo-performance |
| description | Performance specialist for full audits. Measures CWV (LCP/INP/CLS), Lighthouse score signals, and emits deterministic report artifacts. |
Performance Specialist
Use this for the performance sub-track in full audits.
Shared Data Cache
Step 0 -- Check shared data cache:
Before gathering, check .seo-cache/ for cached data from other skills.
Reference: ../seo/references/shared-data-cache.md for schemas and dependency map.
This specialist does not require upstream cache inputs before running fresh performance checks.
- If found: no prerequisite cache is needed; continue with current measurements
- If missing: proceed normally and gather fresh performance data
- If the user says "refresh" or "re-run": ignore cache entirely
Inputs
- URL
- Timeout
- Optional PageSpeed API key (
PAGESPEED_API_KEY)
Execution Path
Prefer running scripts/analyze_performance.py for deterministic performance analysis in API/headless environments. The script uses PageSpeed data when available and otherwise falls back to labeled heuristics so the pipeline still completes without a human in the loop.
Outputs
PERFORMANCE-AUDIT-REPORT.mdSUMMARY.json
Core checks
- Prefer real-user CWV data from PageSpeed Insights / CrUX when available
- If real-user data is unavailable, fall back to Lighthouse lab data and label it clearly as lab data
- Lighthouse performance score (mobile + desktop if available)
- LCP and CLS measurements
- INP only when the data source actually provides it; do not fabricate or infer INP from TBT
- TBT may be reported as a lab proxy when INP is unavailable
