sitemap

name	sitemap
description	Audit an existing sitemap.xml / robots.txt or generate new ones following sitemaps.org protocol and Google Search Central guidance, with a growth-first default that allows every crawler including AI search and AI training bots. Use this skill whenever the user asks to "build a sitemap", "generate sitemap.xml", "audit the sitemap", "refresh sitemap", "write robots.txt", "check robots", "allow AI crawlers", "block AI crawlers", "fix crawl budget", "submit to Google", or any other discovery / indexing configuration task. Output is a valid sitemap.xml (or sitemap index) and/or robots.txt, plus a written audit when auditing.
argument-hint	[audit | build] [optional: site origin]
license	MIT

Overgrow — Sitemap & robots.txt

This skill is the deterministic pipeline for producing and auditing sitemap.xml, sitemap indexes, and robots.txt. It is tuned for growth — by default it invites every compliant crawler, including AI search and AI training bots, unless the user explicitly opts out per user-agent.

Primary reference

Read knowledge$sitemap.md from the plugin root before doing anything. That file is the canonical, rule-by-rule pipeline this skill implements. Follow it exactly. This SKILL.md is the entry-point summary; knowledge$sitemap.md is the source of truth.

Also consult knowledge/geo.md sections on technical SEO and indexability for AI visibility considerations that affect sitemap/robots decisions.

Mode selection

Parse $ARGUMENTS:

• audit (or no args, existing sitemap present) → run the audit flow.
• build (or no args, no sitemap present) → run the generation flow.
• A full origin URL like https://example.com → use as the canonical origin.

If mode is ambiguous, ask the user directly to clarify what you cannot infer.

Audit flow

1. Read .overgrow/inventory.md for the authoritative page list. If missing, run init first.
2. Locate existing sitemap(s): public$sitemap.xml, static$sitemap.xml, sitemap.xml.ts, Next.js app$sitemap.ts, framework-generated routes. If none found, switch to build mode.
3. Locate existing robots.txt: public/robots.txt, static/robots.txt, app/robots.ts, or config-driven.
4. Validate against knowledge$sitemap.md. Report findings:
   • Protocol violations (invalid URL, missing <loc>, bad encoding, over 50k URLs in one file, over 50MB uncompressed).
   • Coverage gaps (pages in inventory missing from sitemap, sitemap URLs missing from inventory — possible 404s or stale entries).
   • lastmod hygiene (missing, future-dated, identical across all entries).
   • Canonical mismatches (sitemap URL vs. rendered canonical tag).
   • robots.txt issues (blocking assets needed for render, blocking sitemap path, contradictory allow/disallow, outdated AI-bot rules).
   • AI-crawler posture (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Bingbot — flag if silently blocked).
5. Write findings to .overgrow/sitemap-audit.md with severity (critical / high / medium / low) and specific fixes.

Build flow

1. Read .overgrow/inventory.md for the list of indexable pages. Exclude utility and anything the user flags.
2. Collect inputs (see knowledge$sitemap.md § "Inputs to collect"). If any input is missing, ask once in a single message.
3. Generate sitemap.xml following the rules in knowledge$sitemap.md. For > 50k URLs, emit a sitemap index plus shard files.
4. Generate or update robots.txt. Default posture: open (allow all user-agents, disallow explicit exclusion paths, reference the sitemap). Only switch to restricted or closed when the user says so — typically for staging, internal, or compliance-locked sites.
5. Write the files to the project's static root (public/, static/, framework-specific path). Do not overwrite existing files without asking; produce .next/.proposed variants if there's an existing file.
6. Print a short summary: URL count, shard count, AI-crawler posture, where the files were written.

Defaults for AI bots (growth-first)

Unless the user says otherwise, robots.txt explicitly allows:

• GPTBot (OpenAI)
• ChatGPT-User
• OAI-SearchBot
• ClaudeBot / anthropic-ai / Claude-Web
• PerplexityBot
• Google-Extended
• Applebot-Extended
• Bingbot (always)
• CCBot (Common Crawl)

Explicit allow lines beat silence — spell them out so the posture is auditable. If the user wants to restrict one or more, ask for the list and write Disallow: / blocks per user-agent.

Constraints

• Never emit sitemap entries for routes that do not exist in the inventory.
• Never include query-stringed URLs unless the user intentionally exposes them.
• Respect canonical tags — the sitemap URL must match the page's own canonical.
• Follow knowledge$sitemap.md for encoding, escaping, lastmod formatting, and sharding rules. Do not improvise.
• Output must pass XML validation against the sitemaps.org schema.

What this skill does NOT do

• Does not submit sitemaps to Search Console / Bing Webmaster — print the URL for the user to submit.
• Does not generate HTML sitemaps (user-facing navigation pages).
• Does not touch hreflang beyond noting issues in the audit — i18n link generation lives in a dedicated sitemap that this skill will build if the inventory declares locales.