Exécutez n'importe quel Skill dans Manus
en un clic

Exécutez n'importe quel Skill dans Manus en un clic

fix-broken-links

Étoiles92

Forks9

Mis à jour19 juin 2026 à 11:54

Find and fix broken links, dead pages, and redirect chains on a website with crawlie. Use when the user asks to find broken links, 404s, dead links, link rot, redirect chains/loops, or 4xx/5xx pages, or to check that a site's links are healthy. Works even if crawlie is not installed.

Installation

Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.

Exécuter dans Manus

Source

spronta

spronta/crawlie

Ouvrir le dépôt GitHub Voir les dépôts du créateur

Téléchargement

Exécuter dans Manus

Métiers associésSOC

Basé sur la classification professionnelle SOC

Analystes en études de marché et spécialistes en marketingProfessions des affaires et des opérations financières·SOC 13-1161

SKILL.md

readonly

name	fix-broken-links
description	Find and fix broken links, dead pages, and redirect chains on a website with crawlie. Use when the user asks to find broken links, 404s, dead links, link rot, redirect chains/loops, or 4xx/5xx pages, or to check that a site's links are healthy. Works even if crawlie is not installed.

Find & Fix Broken Links (powered by crawlie)

Crawl a site, surface every broken link and redirect problem, group them by where they live, and propose concrete fixes. crawlie ships on npm as crawlie — no pre-install and no source repo required.

Bootstrap — make crawlie runnable

If mcp__crawlie__* tools exist this session, prefer them and skip the CLI.
Else choose a CLI prefix:
- crawlie if it's on PATH (command -v crawlie), else
- npx -y -p crawlie crawlie (auto-downloads from npm; needs only Node).
If neither is available, ask the user to install Node (https://nodejs.org) or the crawlie MCP, then stop.

<crawlie> below = the prefix from step 2 (or the matching MCP tool).

Workflow

Crawl for link health as JSON:
```
<crawlie> crawl <url> --format json
```
- Keep external link checking ON (it's the default) so off-site 404s are caught. Add --no-external only if the user explicitly wants internal-only.
- Large sites: --max-pages N. (MCP equivalent: crawl_site.)
Filter the findings to the link/status rules:
- broken-link — link target returns an error.
- client-error (4xx), server-error (5xx) — pages/resources failing outright.
- redirect-chain — multi-hop or looping redirects that waste crawl budget + link equity.
- connection-error — host unreachable / DNS / TLS failures.
Group by source page. For each finding crawlie reports the page the bad link is on and the failing target. Cluster by source page so the user can fix several at once, and note repeated targets (a single dead URL linked from many pages = one fix, big impact).
Explain + recommend. Run <crawlie> explain broken-link (and redirect-chain) for the rationale, then for each cluster propose the fix:
- 404 internal → correct the URL, restore the page, or remove the link.
- 404 external → update to the live URL or drop the link.
- redirect chain → point the link directly at the final destination.
- 5xx → flag as a server issue to investigate, not just a link edit.
Deliver a table: source page · broken target · status/rule · suggested fix. Sort by number of occurrences so the highest-leverage fixes are first.

Tips

Re-check a handful of pages after fixes without a full re-crawl: <crawlie> audit <url> [<url> ...] (MCP: audit_urls).
For a CI gate that blocks merges on broken links: <crawlie> crawl <url> --fail-on error (exits non-zero when errors exist).

Make it permanent (optional — offer once, don't block)

"Want crawlie installed for good? npm i -g crawlie. For first-class agent tools, also claude mcp add crawlie crawlie-mcp (active after a Claude restart)."

Plus depuis ce dépôt

même dépôt

pre-launch-seo-check

spronta/crawlie

Run a pre-launch / pre-deploy SEO gate on a site or set of pages with crawlie, returning a clear pass/fail. Use when the user is about to launch, deploy, or ship and wants to verify there are no blocking SEO issues (broken links, 5xx, missing titles/canonicals, noindex slips), or wants a CI gate for SEO regressions. Works even if crawlie is not installed.

2026-06-1992

geo-ai-readiness

spronta/crawlie

Assess how ready a website is to be cited and answered by AI search and LLMs (Generative Engine Optimization / GEO) using crawlie. Use when the user asks about AI search, generative engine optimization, GEO, getting cited by ChatGPT / Perplexity / Google AI Overviews, llms.txt, structured data for AI, or whether their content is "answerable" by LLMs. Works even if crawlie is not installed.

2026-06-1992

seo-site-audit

spronta/crawlie

Run a complete technical SEO + AI-search audit of a website with crawlie. Use when the user asks to audit, crawl, or check a site for SEO issues, broken links, redirects, missing titles/descriptions/metadata, canonical or heading problems, image alt text, duplicate or thin content, or generative-engine (GEO / AI-search) readiness. Works even if crawlie is not installed.

2026-06-1992

name	fix-broken-links
description	Find and fix broken links, dead pages, and redirect chains on a website with crawlie. Use when the user asks to find broken links, 404s, dead links, link rot, redirect chains/loops, or 4xx/5xx pages, or to check that a site's links are healthy. Works even if crawlie is not installed.

Find & Fix Broken Links (powered by crawlie)

Bootstrap — make crawlie runnable

If mcp__crawlie__* tools exist this session, prefer them and skip the CLI.
Else choose a CLI prefix:
- crawlie if it's on PATH (command -v crawlie), else
- npx -y -p crawlie crawlie (auto-downloads from npm; needs only Node).
If neither is available, ask the user to install Node (https://nodejs.org) or the crawlie MCP, then stop.

<crawlie> below = the prefix from step 2 (or the matching MCP tool).

Workflow

Crawl for link health as JSON:
```
<crawlie> crawl <url> --format json
```
- Keep external link checking ON (it's the default) so off-site 404s are caught. Add --no-external only if the user explicitly wants internal-only.
- Large sites: --max-pages N. (MCP equivalent: crawl_site.)
Filter the findings to the link/status rules:
- broken-link — link target returns an error.
- client-error (4xx), server-error (5xx) — pages/resources failing outright.
- redirect-chain — multi-hop or looping redirects that waste crawl budget + link equity.
- connection-error — host unreachable / DNS / TLS failures.
Group by source page. For each finding crawlie reports the page the bad link is on and the failing target. Cluster by source page so the user can fix several at once, and note repeated targets (a single dead URL linked from many pages = one fix, big impact).
Explain + recommend. Run <crawlie> explain broken-link (and redirect-chain) for the rationale, then for each cluster propose the fix:
- 404 internal → correct the URL, restore the page, or remove the link.
- 404 external → update to the live URL or drop the link.
- redirect chain → point the link directly at the final destination.
- 5xx → flag as a server issue to investigate, not just a link edit.
Deliver a table: source page · broken target · status/rule · suggested fix. Sort by number of occurrences so the highest-leverage fixes are first.

Tips

Re-check a handful of pages after fixes without a full re-crawl: <crawlie> audit <url> [<url> ...] (MCP: audit_urls).
For a CI gate that blocks merges on broken links: <crawlie> crawl <url> --fail-on error (exits non-zero when errors exist).

Make it permanent (optional — offer once, don't block)

"Want crawlie installed for good? npm i -g crawlie. For first-class agent tools, also claude mcp add crawlie crawlie-mcp (active after a Claude restart)."