| name | seo-technical-page-audit |
| description | Use when auditing one webpage for technical SEO, indexability, links, metadata, structured data, or desktop/mobile rendering and performance, then producing a prioritized report with DataForSEO-derived Technical Scores. |
| compatibility | Requires the official DataForSEO MCP server with the ONPAGE module enabled, Python 3, a DataForSEO credential .env file for direct task API access, and filesystem write access. |
SEO Technical Page Audit
Read references/audit-playbook.md before acting and use templates/report-template.md as the report scaffold. Follow its linked official DataForSEO documentation.
Workflow
- Require one absolute HTTP(S) page URL. Derive and normalize its domain; reject malformed or credential-bearing URLs.
- Choose one locale: requested, clear site signal, then
en-US. Select desktop and mobile by default. When the request names only one device, select only it and never call or report the other.
- Prefer MCP task methods. Create one OnPage task per selected device with matching
browser_preset. Run Lighthouse per selected device with full_data: true and for_mobile: false for desktop or true for mobile.
- If task methods are absent, locate the project
.env or ask only for its path. Per selected device run python3 <skill-directory>/scripts/fetch_task_onpage.py '<page-url>' --device <device> with optional --env-file and --accept-language. Accept only exit 0, status: complete, and matching device. Treat exit 2 as a credential-path request. After other failures, use the playbook fallback without automatic billable retries.
- Ask before
on_page_content_parsing_live or on_page_page_screenshot.
- For one device, analyze all evidence in that context. For both, report invariant evidence once: availability, indexability, robots, canonicals, redirects, hreflang, broken links, metadata, headings, content, and schema.
- When both are selected, compare only meaningful render/runtime differences: viewport, touch targets, fonts/scaling, responsive images, resource selection/order and critical path, JavaScript/rendered-content parity, layout shift and performance, throttling, caching/service workers, and mobile-first-indexing risks. Report shared fields by device only when direct evidence materially differs; label these parity issues.
- Claim chains and broken URLs/assets only from exact evidence. Treat resource
checks.is_broken or parser/runtime errors as valid with HTTP 200. Report rounded onpage_score per selected device, or 0 (audit incomplete) when missing. Show dual scores side by side; never blend them or invent weights.
- Use normalized helper fields for device identity and stable summaries, raw payloads for evidence. Prioritize P0-P3; include evidence, impact, fix, owner, effort, and validation.
Cost accounting
Log each endpoint and top-level USD cost. Sum unrounded values. Mark missing-cost subtotals incomplete.
Report file
Use the requested root or <current-working-directory>/SEO, then its domain child. Sanitize <URL> after removing scheme, query, fragment, and trailing slash. Write:
<report-root>/<domain>/<YYYY-MM-DD>_Techical-Report_<URL>.md
Use the local ISO date as the first line. Preserve Techical-Report.
Report structure
Follow the template, remove empty sections, and return the saved absolute path plus a concise summary.
