--- name: growthloud-audit-fix description: Audit the production site behind this repo with Growthloud (SEO + AI-readiness scoring), then fix every failing check at its root in the codebase. Use when asked to improve SEO, AI-search visibility, or to "run a site audit and fix it". --- # Growthloud: audit → fix You are working in the repository of a website. Audit its production deployment, then fix what the audit finds — in code, at the root cause. ## 1. Audit Determine the production URL (check the repo's README, config, or ask). Then: ```bash curl -s -X POST https://app.growthloud.com/api/audit \ -H 'content-type: application/json' \ -d '{"url":"https://YOUR-SITE.com"}' ``` Response shape: ```json { "url": "…", "finishedAt": "…", "scores": { "total": 0, "seo": 0, "ai_readiness": 0 }, "checks": [ { "id": "title", "category": "seo", "weight": 10, "status": "pass|warn|fail|info|na", "evidence": "…", "fix": { "en": "…", "zh": "…" } } ], "shareToken": "…", "cached": false } ``` Quota: 1 fresh audit per IP per day without an account (same-day repeats return the cached result, `cached: true`). A free account at raises the limit to 20/day and adds history, daily re-audit patrol with score-drop alerts, Search Console and Bing panels. ## 2. Fix — every `fail`, then every `warn`, highest `weight` first Find where each check lives **in this repo** (templates, head partials, static files, build config) and fix the root cause — never patch generated output. Follow each check's `fix` text plus this map: | id | where to look / what right looks like | |---|---| | `title` | one ``, ~15–60 chars, names the product and what it does | | `meta-description` | one meta description, ~50–155 chars, concrete value proposition | | `h1` | exactly one `<h1>` per page, matches the page's actual topic | | `canonical` | `<link rel="canonical">` absolute, self-referencing on canonical pages | | `html-lang` | `<html lang="…">` matches the page language | | `hreflang` | bilingual/multilingual sites: reciprocal `alternate` links + `x-default` | | `og-basics` | `og:title`, `og:description`, `og:image` (absolute URL, real asset) | | `structured-data` | valid JSON-LD for the entity (Organization/Product/FAQ as applicable) | | `robots-txt` | `/robots.txt` exists, allows crawling, links the sitemap | | `sitemap` | `/sitemap.xml` exists, lists real canonical URLs | | `search-bots-allowed` | AI *search* crawlers (OAI-SearchBot, Claude-SearchBot, PerplexityBot) not blocked by robots.txt/WAF — this is how AI answers cite you | | `training-bots-policy` | GPTBot/ClaudeBot are *training* crawlers — allowing or blocking them is a policy choice, not a fix; confirm the policy is intentional | | `llms-txt` | `/llms.txt`: plain-markdown facts file — what the product is, pricing bounds, canonical pages; no marketing prose | | `content-extractable` | core content must be present in server-rendered HTML, not injected only by JS | Red lines: no keyword stuffing, no cloaking, no fabricated structured data. Every claim you add to the site must be true of the product today. ## 3. Verify and report Verify each fix on the deployed site (`curl` the page, check the tag is there). The same-day audit is cached, so re-score tomorrow or from a signed-in account. Report: score before, checks fixed (id → change), and the shareable report at `https://app.growthloud.com/r/<shareToken>`.