Codex SEO
Run evidence-led SEO work in Codex. Combine installed specialist skills with a
safe Rust static-analysis core.
Evidence rules
Label every material claim:
- Observed — directly present in fetched HTML, headers, structured data,
supplied exports, screenshots, or tool output.
- Inferred — reasoned from observed evidence. State assumptions.
- Setup required — needs credentials, a connector, a browser session, field
performance data, search results, or another unavailable source.
Never invent rankings, impressions, clicks, backlinks, Core Web Vitals, crawl
coverage, or competitor metrics. Never present static inspection as proof of
indexation or real-user performance.
Compose installed skills
Use $seo as broad orchestrator when installed. Route focused work to matching
$seo-* specialists such as technical, content, schema, sitemap, hreflang,
performance, local, ecommerce, competitor, programmatic, SXO, or GEO.
Use this skill's Rust CLI for deterministic page, sitemap, and drift checks.
Keep specialist reasoning and live-data collection in Codex. Do not install
third-party suites unless user asks.
Choose mode
Read references/audit-playbooks.md only for
requested mode:
- full audit
- single-page review
- technical audit
- content or on-page review
- schema, sitemap, or hreflang review
- local or ecommerce review
- competitor, topic cluster, or programmatic plan
- migration drift
- AI-search or GEO review
Read references/migration-ledger.md when
checking upstream compatibility or deciding whether work belongs in Rust,
specialist skills, or setup-required integrations.
Collect evidence
- Confirm target, scope, locale, audience, business goal, and comparison
baseline when relevant.
- Fetch only URLs user placed in scope. For local files, pass explicit
--base-url when relative links matter.
- Use browser, current official documentation, and connected APIs when live
evidence matters.
- Preserve source URLs and collection time.
- Treat robots directives, canonical tags, schema, and sitemap entries as
signals, not guarantees of search-engine behavior.
Run Rust core
From repository root:
cargo run --quiet --locked --manifest-path skills/codex-seo/scripts/Cargo.toml -- audit --input https://example.com --format markdown
cargo run --quiet --locked --manifest-path skills/codex-seo/scripts/Cargo.toml -- audit --input ./page.html --base-url https://example.com/page --format json
cargo run --quiet --locked --manifest-path skills/codex-seo/scripts/Cargo.toml -- sitemap --input https://example.com/sitemap.xml --format markdown
cargo run --quiet --locked --manifest-path skills/codex-seo/scripts/Cargo.toml -- drift --baseline before.json --current after.json --format markdown
cargo run --quiet --locked --manifest-path skills/codex-seo/scripts/Cargo.toml -- doctor
Installed skill path may differ. Resolve scripts/Cargo.toml relative to this
SKILL.md.
Exit code 0 means command completed without high or critical findings. Exit
code 2 means audit completed and found high or critical findings.
Rust core:
- reads local HTML and XML or public HTTP(S) targets;
- blocks loopback, private, link-local, multicast, documentation, and other
special-use network destinations;
- revalidates each redirect and pins validated DNS results;
- disables ambient proxies and credential forwarding;
- caps redirects, response size, and request time;
- audits titles, descriptions, headings, robots, canonicals, links, images,
language, Open Graph, and JSON-LD;
- validates sitemap structure and URL scope;
- compares JSON audit snapshots for migration drift.
Rust core does not execute JavaScript, authenticate to analytics or search
platforms, crawl arbitrary sites recursively, measure field performance, or
mutate external systems.
Analyze mechanisms
Prioritize causes over symptom counts:
- discovery and crawl path;
- indexability and canonical consistency;
- rendering and content availability;
- intent alignment and information gain;
- internal-link graph and orphan risk;
- structured-data eligibility;
- localization and hreflang graph integrity;
- media delivery and layout stability;
- conversion path and search-experience quality;
- citation-ready passages for AI answer systems.
Explain dependencies. Example: a missing canonical plus conflicting sitemap URL
is one consolidation problem, not two unrelated checklist items.
Return result
Use this structure:
- Scope and evidence — targets, tools, date, inaccessible sources.
- Executive diagnosis — strongest mechanisms and business impact.
- Findings — evidence, severity, confidence, affected URLs, fix, validation.
- Prioritized plan — now, next, later; owner and dependency when known.
- Setup required — exact connector, credential, export, or measurement
needed for unsupported claims.
- Verification — command, re-crawl, browser check, or platform report that
proves each fix.
Keep recommendations specific enough to implement. Include exact tags, URLs,
schema types, internal-link targets, or content changes when evidence supports
them.
Attribution
Workflow adapted from
AgriciDaniel/claude-seo.
See NOTICE.md and LICENSE.txt.
1---2name: codex-seo3description: Audit, plan, and improve technical SEO, on-page content, structured data, sitemaps, hreflang, local or ecommerce visibility, and AI-search citability using evidence-first Codex workflows and a safe Rust static-analysis CLI. Use for SEO audits, page reviews, traffic or ranking investigations, content plans, schema, crawlability, indexation, migration drift, competitor research, and GEO. Never invent search, analytics, backlink, or performance data.4metadata:5 source: https://skillmd.com/skills/decebal/codex-seo6---78# Codex SEO910Run evidence-led SEO work in Codex. Combine installed specialist skills with a11safe Rust static-analysis core.1213## Evidence rules1415Label every material claim:1617- **Observed** — directly present in fetched HTML, headers, structured data,18 supplied exports, screenshots, or tool output.19- **Inferred** — reasoned from observed evidence. State assumptions.20- **Setup required** — needs credentials, a connector, a browser session, field21 performance data, search results, or another unavailable source.2223Never invent rankings, impressions, clicks, backlinks, Core Web Vitals, crawl24coverage, or competitor metrics. Never present static inspection as proof of25indexation or real-user performance.2627## Compose installed skills2829Use `$seo` as broad orchestrator when installed. Route focused work to matching30`$seo-*` specialists such as technical, content, schema, sitemap, hreflang,31performance, local, ecommerce, competitor, programmatic, SXO, or GEO.3233Use this skill's Rust CLI for deterministic page, sitemap, and drift checks.34Keep specialist reasoning and live-data collection in Codex. Do not install35third-party suites unless user asks.3637## Choose mode3839Read [references/audit-playbooks.md](references/audit-playbooks.md) only for40requested mode:4142- full audit43- single-page review44- technical audit45- content or on-page review46- schema, sitemap, or hreflang review47- local or ecommerce review48- competitor, topic cluster, or programmatic plan49- migration drift50- AI-search or GEO review5152Read [references/migration-ledger.md](references/migration-ledger.md) when53checking upstream compatibility or deciding whether work belongs in Rust,54specialist skills, or setup-required integrations.5556## Collect evidence57581. Confirm target, scope, locale, audience, business goal, and comparison59 baseline when relevant.602. Fetch only URLs user placed in scope. For local files, pass explicit61 `--base-url` when relative links matter.623. Use browser, current official documentation, and connected APIs when live63 evidence matters.644. Preserve source URLs and collection time.655. Treat robots directives, canonical tags, schema, and sitemap entries as66 signals, not guarantees of search-engine behavior.6768## Run Rust core6970From repository root:7172```bash73cargo run --quiet --locked --manifest-path skills/codex-seo/scripts/Cargo.toml -- audit --input https://example.com --format markdown74cargo run --quiet --locked --manifest-path skills/codex-seo/scripts/Cargo.toml -- audit --input ./page.html --base-url https://example.com/page --format json75cargo run --quiet --locked --manifest-path skills/codex-seo/scripts/Cargo.toml -- sitemap --input https://example.com/sitemap.xml --format markdown76cargo run --quiet --locked --manifest-path skills/codex-seo/scripts/Cargo.toml -- drift --baseline before.json --current after.json --format markdown77cargo run --quiet --locked --manifest-path skills/codex-seo/scripts/Cargo.toml -- doctor78```7980Installed skill path may differ. Resolve `scripts/Cargo.toml` relative to this81`SKILL.md`.8283Exit code `0` means command completed without high or critical findings. Exit84code `2` means audit completed and found high or critical findings.8586Rust core:8788- reads local HTML and XML or public HTTP(S) targets;89- blocks loopback, private, link-local, multicast, documentation, and other90 special-use network destinations;91- revalidates each redirect and pins validated DNS results;92- disables ambient proxies and credential forwarding;93- caps redirects, response size, and request time;94- audits titles, descriptions, headings, robots, canonicals, links, images,95 language, Open Graph, and JSON-LD;96- validates sitemap structure and URL scope;97- compares JSON audit snapshots for migration drift.9899Rust core does not execute JavaScript, authenticate to analytics or search100platforms, crawl arbitrary sites recursively, measure field performance, or101mutate external systems.102103## Analyze mechanisms104105Prioritize causes over symptom counts:106107- discovery and crawl path;108- indexability and canonical consistency;109- rendering and content availability;110- intent alignment and information gain;111- internal-link graph and orphan risk;112- structured-data eligibility;113- localization and hreflang graph integrity;114- media delivery and layout stability;115- conversion path and search-experience quality;116- citation-ready passages for AI answer systems.117118Explain dependencies. Example: a missing canonical plus conflicting sitemap URL119is one consolidation problem, not two unrelated checklist items.120121## Return result122123Use this structure:1241251. **Scope and evidence** — targets, tools, date, inaccessible sources.1262. **Executive diagnosis** — strongest mechanisms and business impact.1273. **Findings** — evidence, severity, confidence, affected URLs, fix, validation.1284. **Prioritized plan** — now, next, later; owner and dependency when known.1295. **Setup required** — exact connector, credential, export, or measurement130 needed for unsupported claims.1316. **Verification** — command, re-crawl, browser check, or platform report that132 proves each fix.133134Keep recommendations specific enough to implement. Include exact tags, URLs,135schema types, internal-link targets, or content changes when evidence supports136them.137138## Attribution139140Workflow adapted from141[AgriciDaniel/claude-seo](https://github.com/AgriciDaniel/claude-seo).142See [NOTICE.md](NOTICE.md) and [LICENSE.txt](LICENSE.txt).