15 lines
2.1 KiB
Plaintext
15 lines
2.1 KiB
Plaintext
GOAL - W5b Truly unique titles + meta descriptions per page (odoo-expertos). Branch: fix/w5b-unique-meta.
|
|
Repo: C:\Users\eugen\.cursor\projects\Odoo Expertos\odoo-expertos. Work ONLY in this repo.
|
|
Context: The W5 codemod appended slug+lang as a meta-suffix to dedupe, which produces ugly truncated SERP previews like "...cuando elegir cad - Odoo Vs Sap Hosting 2026 | ES" and still leaves 194 duplicate titles. AI engines and Google reward genuinely-unique, content-derived metadata.
|
|
TASK: Write ONE codemod (Node or Python) that for every <slug>/index.html:
|
|
(a) Extracts the page's H1, its first H2, and the first 2 sentences of body prose (skip nav/footer/script/style).
|
|
(b) Generates a UNIQUE <title> tag of 50-60 characters from {H1-core-phrase} - {value-fragment derived from first H2} | Odoo Expertos. If the produced title would clash with any other page's title, append the most distinctive noun phrase from the first H2.
|
|
(c) Generates a UNIQUE <meta name="description"> of 140-160 characters from the first 2 sentences, lightly cleaned (no run-on, ends with action verb where natural), in the page's declared language. Do NOT use the slug+lang suffix trick.
|
|
(d) After rewriting, verify uniqueness corpus-wide: no two pages share a title; no two pages share a meta description.
|
|
CONSTRAINTS: only edit <title>, <meta name="description">, and (if present) <meta property="og:title"> / <meta property="og:description"> to mirror. Do not touch body content. Preserve language per page.
|
|
PROVE each turn: run powershell -File seo-audit-2026-05-27\REVIEW.ps1 -Only W5b and paste full output.
|
|
DONE WHEN: that output contains "W5b: PASS" (duplicate titles=0 AND duplicate metas=0 AND all titles 40-65 chars AND all metas 130-165 chars), AND committed on branch fix/w5b-unique-meta (show git log -1 --oneline). Stop after 25 turns if blocked.
|
|
|
|
|
|
CONCRETE TARGETS: read seo-audit-2026-05-27\OPPORTUNITIES-2026-05-28.md section 3 for the exact content-derived title+meta generator pseudocode and 5 worked examples. Implement that rule (extract H1 + first H2 + first 2 body sentences; skip nav/header/breadcrumb when scanning paragraphs; uniqueness registry across the corpus).
|