Skip to content

Example workflows

Scenario-based conversation starters for the 96-tool registry in release 1.6.0. Each file covers one use case with a progression from a first question to a deeper investigation.

Start with the first prompt in any file and replace yourdomain.com with your property or verified-site URL. Configure the required providers with the installation guide, Google setup or Bing setup.

For a measured example, read the Claude Code Ultimate Guide run, recorded on 2026-10-07. It includes Google metrics, page checks, suggested actions and a shareable MCP trace. No site changes or ranking improvements are claimed.

FileProvidersUse caseDepth
quick-audit.mdGoogle, optional BingSite health check at a glanceQuick
google-bing-comparison.mdGoogle + BingCompare engines without mixing incompatible metricsIntermediate
full-audit.mdGoogle, optional Bing and GA4Complete audit with P0/P1/P2 action planComplete
full-audit-to-fixes.mdAll relevant available tools + GitHubFrench prompt: audit, publish epic and issues, then implement and validate fixesComplete + implementation
keyword-opportunities.mdGoogle + supported Bing analysesFind measured ranking opportunitiesIntermediate
page-deep-dive.mdGoogle, optional Bing, CrUXDiagnose a single URLIntermediate
traffic-drop.mdGoogleInvestigate a traffic drop across exact adjacent periodsIntermediate
indexing-issues.mdGoogleInspect indexing evidence and eligible submissionsIntermediate
content-brief.mdGoogle, optional GA4 and CrUXBuild a brief from observed search dataIntermediate

For the new tools, copy the 1.4.0 smoke test, weekday reference, declared-change and crawl-preview prompts. Read draft/rewrite workflows and bounded audit workflows for their coverage limits.

Copy a prompt into Claude, Codex or another MCP-compatible client. Scenario files contain starting questions, evidence-preserving follow-ups and explicit boundaries for writes. The audit-to-fixes prompt is a complete project instruction that authorizes GitHub issues and repository corrections after evidence collection.

The prompts work as-is. You don’t need to know which API tool runs behind each question.

Read-only analysis does not authorize a write. Before a sitemap, URL, feed or IndexNow mutation, the assistant must name the exact tool and target, report the number of affected items, and wait for explicit confirmation. An accepted submission proves neither crawl nor indexation.

The 1.5.0 local smoke test, indexing matrix and crawl snapshot prompts cover the new local evidence tools without requiring provider access.