Agent Skills: SEO audit report (scaffold instructions)

SEO audit report (scaffold instructions)

Use this skill when the user wants an interactive, client-side report for inspecting Search Console exports or connected data repeatedly in the browser.

Context and preflight

Run the shared preflight in ../docs/credentials-and-tooling.md before discovery or implementation. Reuse .seo-context.md when it exists. Resolve the project's canonical GSC property from accessible URL-prefix and sc-domain: variants. Record unresolved assumptions rather than inventing business priorities or edit permission.

Required inputs

• Project directory name.
• Data source: live connected GSC data or an export created from the connected MCP.
• Persistence target: a serialized sql.js database persisted in the browser's IndexedDB.

Tooling & credentials

• Auth mode: mcp
• Requires: connected GSC access through the required MCP
• Input path: live GSC data or an MCP export materialized to JSON/CSV
• Fallback: none

Workflow

1. Scaffold the project

   • Run npm create vite@latest <project-name> -- --template react-ts (or bun create vite@latest ... if Bun is preferred).
   • Enter the new folder and install deps (npm install/bun install).
   • Keep the generated project structure unless the requested report needs a different route or entry point.

2. Add dependencies

   • Install react, react-dom (if not already) along with zod.
   • Install runtime helpers: sql.js for the in-browser SQLite engine and idb-keyval for persisting exported database bytes.
   • Keep typescript + vite dev dependencies aligned with the template.

3. Implement the data layer

   • Configure sql.js with the sql-wasm.wasm bundle and isolate initialization in a reusable helper.
   • Parse CSV and JSON inputs into { date, query, page, clicks, impressions, ctr, position }, coercing numeric fields and handling missing optional values consistently.
   • Store normalized rows in a gsc_rows table with indexes on date, query, and page.
   • Export the database to bytes after imports and persist those bytes with idb-keyval; restore them when the app loads.

4. Implement the report views

   • Surface total clicks, impressions, CTR, and weighted position, plus top queries, top pages, and CTR-gap opportunities.
   • Make opportunity thresholds and date/page/query filters visible and adjustable rather than hiding them in query code.
   • Include file import, optional fetch from /gsc/latest.json, custom filters, and reset controls.

5. Polish and document the scaffold

   • Style the layout with CSS variables and card/grid patterns, keeping readability front-and-center with a sticky header, callouts, and a narrow text width.
   • Add a small sample export such as public/gsc/latest.json when it helps demonstrate the flow.
   • Document the accepted import schema, persistence model, and CTR-opportunity heuristics in the README or nearby comments.

6. Run and validate

   • Start the dev server with npm run dev or bun run dev and load a representative CSV/JSON export.
   • Exercise import, reload persistence, filters, reset, and the optional /gsc/latest.json fetch path. Check the browser console for worker/WASM or IndexedDB errors.
   • Run the project's available typecheck, lint, and build commands. In the README, state which commands ran and any live-GSC or browser limitations.
   • Remember that sql.js executes SQLite in the browser; IndexedDB stores the serialized database bytes—it is not the SQLite engine.