Read-only SEO HTTP audit with saved evidence

Fetch pages, robots.txt and sitemaps with GET only, keep the raw evidence, and report titles, canonicals, noindex and image crawlability with proof.

Claude Hermes Agent Codex CursorMIT (modify, redistribute and use commercially)Python

What it does

Audits a live site's technical SEO without changing anything.

Fetches pages, robots.txt and sitemaps (index to child sitemaps, recursively) and saves bodies, headers, redirects, hashes and timestamps
Extracts title, meta description, canonical, meta robots, X-Robots-Tag, H1s, image alt text and OG images
Evaluates robots.txt with Google's documented rules (longest match, Allow wins ties) and checks whether images are actually crawlable
Classifies findings as observed defects, items needing Search Console, or things that are not automatically defects

How to use

python scripts/seo_audit.py https://www.example.com/ --out ./seo-audit

It writes seo-audit/report.json and evidence/. The agent then reports priorities, target URLs, evidence and remedies, citing Google's official documentation.

Requirements

Python 3 (standard library only)

Notes

No publishing, CMS edits, Search Console registration or URL submission during the audit.
Indexing and ranking are never judged without Search Console data.
Keep concurrency and delays conservative to avoid loading the target site.

How to install

Unzip the downloaded file (it contains a readonly-seo-http-audit/ folder with SKILL.md).
Copy the whole folder into your agent's skills folder.
- Claude: ~/.claude/skills/ (per project: .claude/skills/)
- Hermes Agent: ~/.hermes/skills/
- Codex: ~/.agents/skills/ / Cursor: ~/.cursor/skills/
- Claude app (claude.ai): upload the ZIP in Customize → Skills
Restart the agent and check that the skill appears in its skill list.

Step-by-step install for Windows / Mac and each agent