What does this skill do, and when should you use it?
SEO Drift is an Agent Skill focused on page-level SEO regression monitoring. It captures a known-good snapshot, compares later page states, and records the change history. Tracked data includes titles, meta descriptions, canonicals, robots directives, headings, JSON-LD, Open Graph, Core Web Vitals, HTTP status, and content hashes. It fits deployment checks and traffic-drop investigations, but it is not a replacement for a full SEO audit.
It fetches pages through an SSRF-protected pipeline, parses SEO-critical HTML elements, and can query PageSpeed for Core Web Vitals. Baselines and comparison results are stored locally in SQLite. The comparison engine applies 17 rules, classifies findings as CRITICAL, WARNING, or INFO, and outputs JSON containing triggered rules, old and new values, severity, and recommended actions. It also shows baseline and comparison history for a URL and can generate an HTML report from a comparison result.
- An SEO or engineering team captures a baseline before deployment and runs compare afterward to check titles, canonicals, schema, and status codes for regressions.
- A site owner investigating an organic-traffic drop uses compare and history to identify what changed and when.
- A publishing team periodically baselines important pages to detect accidental changes to Core Web Vitals, Open Graph tags, or heading structure.
- A technical SEO team uses drift findings to route schema, performance, page, content, or technical follow-up work to the appropriate specialized skill.
How do you install this skill?
- The referenced seo-drift scripts and drift-specific tests were not supplied, so command executability cannot be confirmed statically.
- The optional CWV step depends on an external PageSpeed service; network access, API configuration, and URL data disclosure should be verified.
- The database uses a fixed ~/.cache/claude-seo/drift/baselines.db location without documented backup, migration, cleanup, or recovery procedures.
- Several SEO-impact and response-time statements are empirical claims without direct supporting sources in the skill.
- Shell / CLI
- Network access
- Local filesystem
Python 3.10+SQLite
The skill is located at skills/seo-drift/SKILL.md in the repository. The README documents installation of the full Claude SEO collection: for the plugin route, run /plugin marketplace add AgriciDaniel/claude-seo and /plugin install claude-seo@agricidaniel-claude-seo; on Unix/macOS/Linux, run git clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git followed by bash claude-seo/install.sh. The source does not document a separate seo-drift-only installation procedure.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/seo-drift" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
In Claude Code, run /seo drift baseline <url> to capture a baseline, /seo drift compare <url> to compare the current page, and /seo drift history <url> to review history. Use --skip-cwv to skip Core Web Vitals, and use --baseline-id with compare to select a specific baseline. Data is stored by default at ~/.cache/claude-seo/drift/baselines.db.
What are this skill's strengths and limitations?
- The monitoring scope is concrete, covering metadata, structured data, performance, status codes, and content hashes.
- Local SQLite storage supports repeatable baselines and historical review.
- Comparison output includes severity, old and new values, and recommended actions.
- The workflow includes URL validation, SSRF protection, parameterized SQL queries, and TLS verification.
- It needs network access to fetch target pages; Core Web Vitals checks may also require PageSpeed API access.
- It depends on repository scripts and a local SQLite database, so the SKILL.md alone is not a standalone implementation.
- The source does not document a platform-neutral installation flow; Codex or other clients would need command-invocation adaptation.
- Its scope is regression detection, not a full site audit, keyword research, or backlink analysis.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| SEO Drift Monitor this page | 49 · Use with care | ★ 19k | 23d ago | MIT |
| Full-Site SEO Audit | 53 · Use with care | ★ 19k | 23d ago | MIT |
| Single-Page SEO Inspector | 51 · Use with care | ★ 19k | 23d ago | MIT |
| SEO Content Audit Skill | 48 · Use with care | ★ 2.8k | 8mo ago | MIT |
| Claude SEO Site Analysis | 46 · Use with care | ★ 19k | 23d ago | MIT |
Compared with manual before-and-after checks, it provides structured baselines, fixed comparison rules, and retained history. Compared with a full SEO audit tool, it is narrower: its main purpose is detecting regressions in known page states rather than finding every SEO issue.
How did FollowSkills review this skill?
The skill explicitly describes local SQLite storage, SSRF validation, parameterized queries, and TLS verification, with no observed malicious or destructive default. It lacks explicit user confirmation, disclosure of external PageSpeed data flows, database backup/rollback, and finer-grained storage isolation, so points are deducted.
The workflow, tracked fields, 17 rules, and abnormal-input handling are fairly specific, and the repository provides CI and test evidence. However, the supplied material does not include the skill scripts or drift-specific tests covering the key paths, so runtime consistency and diagnosable failures cannot be confirmed; the score remains conservative under the static cap.
Trigger phrases, URL arguments, baseline/compare/history scenarios, and JSON outputs are clear. Non-fit boundaries, Chinese-language behavior, and mainland-China reachability are not documented, while CWV may depend on an external PageSpeed service and credentials, so points are deducted.
The skill has readable structure, an argument hint, MIT licensing, author and 2.2.0 metadata, and practical workflows. It lacks skill-specific installation and dependency notes, FAQs, maintenance ownership, update path, changelog, and deeper troubleshooting guidance, so it does not receive full marks.
Baseline, diff, and history goals are concrete, and the rules define severity, thresholds, old/new values, and actions that could support deployment regression checks. No directly reusable output or statically verifiable key-path result is supplied, and CWV failures reduce coverage, so the score is limited by static calibration.
The comparison rules, field-source table, and security assertions provide some auditability, and repository CI adds limited context. The supplied evidence lacks the drift scripts, drift-specific tests, and independent corroboration needed to reproduce or validate the core conclusions, so only limited static evidence credit is given.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →