← All skills

🧰 Internal Link Suggester

Reads a folder of Markdown blog posts and suggests where a new or existing post should link to related posts, with the exact sentence and anchor text for each link.

SEO/GEO writingIntermediate · Edit variables or configuration before use.FreePublished Nov 1, 2026
Tested on✓ CodexSee test log
Free download

ZIP with SKILL.md and any reference files.

Download ZIP

What it does

Point the skill at your content folder and name one target post. It reads titles, headings and body text, picks the posts that are actually on the same subject, and proposes links in both directions: from the target out to related posts, and from related posts back to the target. It suggests edits; it only writes them if you say so.

Use it when: you just finished a post and want it woven into the rest of your blog without hunting through old articles by hand.

Works with: Claude Code, Cursor and Codex, on any static-site repo with Markdown or MDX posts (Astro, Hugo, Next.js, Jekyll).

Output preview

Task given to the agent: “Read and follow .agents/skills/internal-link-suggester/SKILL.md before you answer. Use that skill's output format. Suggest internal links for the post content/posts/packing-list-basics.md using only the posts in content/posts/. Max 4 outbound and 4 inbound. Do not edit files; report suggestions in the skill's table format. URL pattern is /blog/<slug>/.”

AFTER · with the skill
Target: Packing List Basics for Importers (/blog/packing-list-basics/)

Outbound links
| # | Sentence (quoted) | Anchor text | Link to |
|---|---|---|---|
| 1 | "A packing list tells the warehouse what is in each carton." | packing list | /blog/packing-list-basics/ |
| 2 | "It should match the commercial invoice line by line." | commercial invoice | /blog/commercial-invoice-checklist/ |
| 3 | "When carton counts disagree with the invoice, the consignee spends hours at the dock." | carton counts disagree | /blog/commercial-invoice-checklist/ |
| 4 | "Missing item codes and vague descriptions cause most mismatches." | item codes and vague descriptions | /blog/commercial-invoice-checklist/ |

Inbound links
| # | From post | Sentence (quoted) | Anchor text |
|---|---|---|---|
| 1 | Commercial Invoice Checklist | "Each line needs SKU, description, quantity, unit price and amount." | packing list basics |
| 2 | Commercial Invoice Checklist | "Totals must add up." | packing list |
| 3 | Factory Audit Overview | "A factory audit looks at quality systems, social compliance and capacity." | packing list basics |
| 4 | Factory Audit Overview | "It is not a packing list check." | packing list check |

Skipped: none

Real, unedited output from recorded runs on Nov 1, 2026 · Codex (codex-cli 0.163.0) · model gpt-5.4-mini. Test inputs are fixtures (fictional brand, example.com links). AI output can be wrong: check facts before you publish anything like it.

Install

Unzip first. You get a folder named internal-link-suggester containing SKILL.md.

Claude Code

Put the folder in your personal skills directory, or in the project to share it with your team.

~/.claude/skills/internal-link-suggester/SKILL.md.claude/skills/internal-link-suggester/SKILL.md

Claude Code picks up new skills in the current session (run /reload-skills if the skills folder didn't exist when the session started). The skill loads when a task matches its description.

Claude.ai

Turn on code execution (Settings → Capabilities), then go to Customize → Skills, click +, then Create skill → Upload a skill, and select the ZIP. Full guide.

Cursor

Put the folder in your user or project skills directory. Cursor also reads .claude/skills.

~/.cursor/skills/internal-link-suggester/SKILL.md.cursor/skills/internal-link-suggester/SKILL.md

Codex

Put the folder in your user or repository skills directory, then restart Codex or run /skills to check it is listed.

~/.agents/skills/internal-link-suggester/SKILL.md.agents/skills/internal-link-suggester/SKILL.md

These are the standard skill locations for each tool. The tools a skill is verified on are listed under “Tested on” above, if any.

Test log

Date (UTC)ToolModelRunResultChecksTokens in / out
Nov 1, 2026Codexgpt-5.4-miniTestPassed
4/4
  • ✓ codex exited cleanly
  • ✓ skill was loaded (SKILL.md read)
  • ✓ output has substance (80+ words) (201 words)
  • ✓ no obvious secret leak
38,390 / 600
Nov 1, 2026Codexgpt-5.4-miniTestPassed
4/4
  • ✓ codex exited cleanly
  • ✓ skill was loaded (SKILL.md read)
  • ✓ output has substance (80+ words) (203 words)
  • ✓ no obvious secret leak
50,070 / 853
Nov 1, 2026Codexgpt-5.4-miniTestFailed
3/4
  • ✓ codex exited cleanly
  • ✓ skill was loaded (SKILL.md read)
  • ✗ output has substance (80+ words) (51 words)
  • ✓ no obvious secret leak
49,460 / 386

Each row is a real run recorded by our test runner. For each tool, the latest test run decides whether it appears under “Tested on”. Baseline runs do the same task with no skill installed and never count.