# Aie Web Links

> Website Link Checker

- Skill: `mrhinkle/aie-web-links` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add mrhinkle/aie-web-links`
- Raw SKILL.md: https://api.skillmd.com/api/skills/mrhinkle/aie-web-links/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: mrhinkle (https://skillmd.com/u/mrhinkle)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/mrhinkle/aie-web-links

---


# Website Link Checker

Crawls a site and validates links, images, and assets, then reports what's broken and how to fix it. Broken links hurt user trust and SEO; this catches them before visitors do.

## Workflow

1. **Confirm the source** — a local build/source folder OR a publicly reachable URL (production or staging). Note any auth or robots.txt restrictions; report links that can't be reached rather than guessing.
2. **Build the link inventory** — parse pages for `<a href>`, `<img src>`, `<link href>`, `<script src>`, media/`srcset`, `mailto:`, `tel:`, and in-page `#anchor` targets. Note where each link lives (source page + line).
3. **Validate each link** (see `reference/status-codes.md` for how to interpret responses):
   - Internal links resolve to a real page/asset.
   - External links return a healthy status; distinguish broken (404/410/5xx) from merely slow.
   - Images/assets load.
   - No HTTP asset is referenced on an HTTPS page (mixed content).
   - Redirects are single hops, not chains.
   - `#anchor` links point to an element with a matching `id`/`name` on the target page.
   - `mailto:`/`tel:` are well-formed.
4. **Find orphans** — pages that exist but nothing links to them (unreachable by crawl from the homepage).
5. **Report** using the format below. See `reference/fixes.md` for the standard fix for each issue type.

## Practical notes

- Some external sites block automated requests (403) or rate-limit. Report these as "could not verify," not confirmed-broken.
- A 200 with a slow response is a warning, not a break.
- Prefer `HEAD` requests where supported, falling back to `GET`, to reduce load.
- For large sites, cap external-link concurrency and de-duplicate URLs before requesting.

## Output Format

Lead with a summary, then findings grouped by severity.

**Summary**
- Links checked: N (internal / external / assets)
- Broken: N • Redirect chains: N • Mixed content: N • Orphans: N
- Quick wins (fix in under 5 min): N

**Findings** — for each issue:
| Severity | Source page | Link / target | Status | Fix |
|----------|-------------|---------------|--------|-----|

Severity levels:
- **Critical** — broken internal links, mixed content on HTTPS, links to pages returning 5xx, missing images on key pages (home/pricing/CTA).
- **High** — broken external links on high-traffic pages, invalid email links, redirect chains, non-working social links.
- **Medium** — missing images on secondary pages, orphaned pages, broken anchor links, slow-but-working links.
- **Low** — outbound links to deprecated-but-working pages, slow external assets.

If URLs changed, also output a **redirect map** (old URL → new URL → 301/302).

## Reference

- Read `reference/status-codes.md` for the status-code table and what each means.
- Read `reference/fixes.md` for the standard remediation and time estimate per issue type.

## Standards

- Google Search Central on crawling and links — https://developers.google.com/search/docs/crawling-indexing
- MDN Mixed content — https://developer.mozilla.org/en-US/docs/Web/Security/Mixed_content
- MDN HTTP redirects — https://developer.mozilla.org/en-US/docs/Web/HTTP/Redirections