Jessie confirms the task is a technical audit and identifies one public website.
Live Jessie skill
Technical SEO Audit.
Runs a public-site crawl audit through the execution plane, using live MCP evidence, optional Search Console enrichment, and the deterministic technical_seo_analyze MCP tool to produce the report package.
When It Routes
A task asks for a technical SEO audit, site crawl audit, Screaming Frog-style audit, indexability review, canonical review, or crawl analysis.
Flow
The run receives execution-plane MCP grants for Browser Run crawl polling, PageSpeed sampling, the technical_seo_analyze package builder, SEO context tools, /work/output/ artifact persistence, and Asana write-back delivery.
The execution plane records sandbox_created, workspace_hydrated, mcp_tools_listed, and harness_started in D1, while mcp-tools records first_tool_call and crawl_submitted with the Browser Run job id.
Jessie loads approved site/context notes, known migrations, platform constraints, and consultant instructions.
Jessie crawls the public site, checks deterministic technical SEO signals, samples PageSpeed, optionally enriches with per-client Search Console metrics and URL Inspection evidence, and stores evidence files.
The technical_seo_analyze MCP tool builds the evidence-backed findings, report, manifest, and remediation companion from the gathered JSON inputs.
Jessie attaches the execution-plane report artifacts and posts a concise internal review comment.
Context Jessie Looks At
- Approved client/site context, known migrations, CMS constraints, current technical priorities, and consultant notes.
- Task-specific target URL, scope, and questions.
- Recent strategy/report context if the task links to it.
Data And Tools
- Cloudflare Browser Run crawl records.
- Robots.txt and XML sitemap fetches.
- Sampled PageSpeed Insights evidence.
- Targeted rendered validation when JavaScript SEO risk needs checking.
- Optional Google Search Console Search Analytics page metrics and URL Inspection results when a per-client Source Asset maps a site URL.
- Execution-plane /work/output/ report, manifest, evidence files, and Asana attachments.
LLM Insight Stage
The execution-plane agent gathers live evidence, then calls the deterministic technical_seo_analyze MCP tool. Optional Search Console maps are deterministic analyzer inputs. The analyzer remains the source of truth for findings, evidence files, and the consultant report package; the agent must not invent findings outside that package.
Outputs
- Technical SEO audit HTML report with an executive decision summary, plain-language interpretation, priority roadmap, target state, evidence package manifest, technical coverage inventory, and workstream action briefs.
- Action briefs state decision owner, implementation owner, likely control surface, decision needed, why it matters, target good state, direct acceptance checks, validation steps, and evidence.
- Evidence package with crawl data, page inventory, internal-link source evidence, debug data, issue CSVs, support exports, manifest, and remediation companion.
- Asana summary comment with a decision summary, priority roadmap, top recommended checks, and package references.
Guardrails
- Public websites only; no authenticated, staging, private, CAPTCHA, or login areas.
- No robots.txt bypass, CMS changes, tracking changes, or follow-up task creation in v1.
- An empty or incomplete crawl is reported as a crawl-quality issue, not success.
- No completed audit when crawl evidence, PageSpeed evidence, site-discovery evidence, rendered-validation evidence, or analyzer outputs are missing or synthetic.
- PageSpeed findings must name or commission discovery of the exact LCP/resource, render-blocking asset, unused bundle, image/font issue, or third-party cause before implementation is assigned.
- Search Console enrichment is optional. Missing access, missing per-client site URL, or URL Inspection quota limits add a caveat and must not block report rendering.
Example Prompt
Run a technical SEO audit for https://www.example.com.au.