Catching an SEO regression after it has already cost traffic is expensive.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/AgricIDaniel/claude-seo/llms.txt
Use this file to discover all available pages before exploring further.
/seo drift solves this by treating your SEO state the same way Git treats code: capture a baseline, diff it later, and review the full history whenever you need to explain what changed. The command was contributed by Dan Colta (Pro Hub Challenge, v1.9.0). The companion command /seo flow integrates the FLOW evidence-led SEO framework (CC BY 4.0) — 41 structured AI prompts across five campaign stages — so every analysis decision is grounded in primary-source evidence rather than improvised queries.
/seo drift baseline|compare|history <url>
/seo drift stores all snapshots in a local SQLite database at ~/.cache/claude-seo/drift/baselines.db. No external service is required. URL normalization ensures consistent matching across runs: lowercase scheme/host, strip default ports, sort query parameters, remove UTM parameters, strip trailing slashes.
What gets snapshotted
Every baseline records 13 SEO-critical elements:| Element | Field | Source |
|---|---|---|
| Title tag | title | parse_html.py |
| Meta description | meta_description | parse_html.py |
| Canonical URL | canonical | parse_html.py |
| Robots directives | meta_robots | parse_html.py |
| H1 / H2 / H3 headings | h1, h2, h3 (arrays) | parse_html.py |
| JSON-LD schema | schema (array) | parse_html.py |
| Open Graph tags | open_graph (dict) | parse_html.py |
| Core Web Vitals | cwv (dict) | pagespeed_check.py |
| HTTP status code | status_code | fetch_page.py |
| HTML content hash | html_hash (SHA-256) | Computed |
| Schema content hash | schema_hash (SHA-256) | Computed |
--skip-cwv to capture all other fields without it.
17 comparison rules across 3 severity levels
The comparison engine applies 17 rules when diffing a current page state against a stored baseline:| Severity | Meaning | Response time |
|---|---|---|
| CRITICAL | SEO-breaking change, likely traffic loss | Immediate |
| WARNING | Potential impact, needs investigation | Within 1 week |
| INFO | Awareness only, may be intentional | Review at convenience |
noindex added to robots directives, title tag removed or emptied, HTTP status changed to a 4xx/5xx error. Examples of WARNING findings: title tag modified, H1 changed, schema type changed or removed.
Typical drift workflow
Capture the initial baseline
Run this before a deployment, migration, or major content update — or simply on a Monday morning to establish the week’s reference point.The command outputs a baseline ID, timestamp, URL, and a summary of all captured elements.
Run a weekly comparison
After a deployment or after a week has passed, diff the current page state against the stored baseline.The output lists all triggered comparison rules with old value, new value, severity classification, and a recommended action for each finding.
Generate an HTML drift report
After any comparison, generate a shareable HTML report for client or team communication.
Review the history
When a traffic drop appears in Search Console, check the drift history to pinpoint when and what changed.Returns all baselines and comparisons newest-first, with timestamps and per-comparison summaries.
Hand off to the right specialist command
Drift detects the regression; specialist commands diagnose the root cause. The comparison output includes a cross-skill recommendation for every finding type: schema removal →
/seo schema, CWV regression → /seo technical, canonical change → /seo technical, noindex added → /seo technical, H1 changed → /seo content.Use case: moving from snapshots to “what changed this week”
The key value of drift monitoring for agency workflows is shifting client conversations away from point-in-time audit PDFs toward a continuous regression signal. An SEO agency running 10 client sites can baseline each site on Monday, run a compare on Friday, and send clients a diff-based health update in 12 minutes rather than rebuilding a full audit each week. Regressions between audits — a canonical quietly breaking after a CMS update, a schema type disappearing after a template change — are caught before the next ranking data drop shows up in GSC./seo flow [stage]
/seo flow integrates the FLOW framework (© Daniel Agrici, CC BY 4.0) — an evidence-led SEO operating model built for the AI-search era. The framework’s 41 AI prompts are organized into five campaign stages. Every prompt is grounded in a specific evidence source (SERP data, GSC, GA4, customer interviews) with attribution preserved. The framework is cited in the plugin’s Ecosystem section as the knowledge base powering all seo-flow prompts.
The five FLOW stages
FIND — 5 prompts
FIND — 5 prompts
The Find stage covers keyword research, SERP intent mapping, and content gap analysis. Prompts are driven by actual SERP signals — related searches, People Also Ask clusters, and intent classification — rather than seed-keyword guessing. Cross-reference:
/seo cluster <seed-keyword> complements the Find stage with full SERP-overlap semantic clustering.LEVERAGE — 1 prompt
LEVERAGE — 1 prompt
The Leverage stage focuses on backlink strategy and off-site authority building: identifying link gaps, prioritizing outreach targets, and positioning for “best of” list placements (the top AI visibility citation factor per Whitespark 2026). Cross-reference:
/seo backlinks <url> provides the raw backlink data that the Leverage prompt acts on.OPTIMIZE — 21 prompts
OPTIMIZE — 21 prompts
The Optimize stage is the largest: 21 prompts covering on-page SEO, E-E-A-T signals, content freshness, schema markup, Core Web Vitals, and technical issues. The skill selects exactly 2–3 prompts per run based on industry vertical, prior skill output in the conversation, and URL signals — never dumps all 21. Cross-reference:
/seo content <url> and /seo geo <url> for full content quality and AI-search readiness analysis.WIN — 3 prompts
WIN — 3 prompts
The Win stage targets bottom-of-funnel pages: conversion rate optimization, dual-surface SERP scoring (organic + AI Overviews), and BOFU content strategy. Cross-reference:
/seo sxo <url> for persona-based scoring of landing page and product page alignment with SERP intent.LOCAL — 11 prompts
LOCAL — 11 prompts
The Local stage is the most populated: 11 prompts covering GBP optimization, meta descriptions, title tags for local intent, citation strategy, and structured local audit workflows. Cross-reference:
/seo local <url> and /seo maps [command] for full local SEO and geo-grid rank tracking.How FLOW connects to other /seo commands
The FLOW stages map directly onto the Claude SEO command set. A typical campaign run might look like:
/seo flow activation outputs the framework attribution before analysis: Framework and prompts © Daniel Agrici, CC BY 4.0 — github.com/AgriciDaniel/flow.