agentleFS
adityash8adityash8/claude-seo-skills/.claude/skills/internal-linking/SKILL.md

internal-linking

Skill0 starsChanged 21 days agoMITView on GitHub

Use when planning, adding, auditing, or measuring contextual internal links, orphan recovery, hubs, or anchor rewrites.

8 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. 0 are common, 15 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 8 edited versions we read keep these.

  • 15

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Search or analytics data when ranking donors or measuring results.

    line 25
  2. Distinctive:

    Never change a published source to Draft to stage a link.

    line 34
  3. Distinctive:

    Keep one planned body edge per source and target.

    line 36
  4. Very specific:

    Use ordinary crawlable links.

    line 37
  5. Distinctive:

    Verify the target returns a direct HTTP 200, is indexable, and has the intended canonical.

    line 38
  6. Distinctive:

    Start with a bounded manual batch.

    line 41
  7. Distinctive:

    Use evidence gathered after relevant canonical, redirect, or indexability fixes.

    line 42
  8. Distinctive:

    Read back the saved content and publication status.

    line 43
  9. Distinctive:

    Use concise descriptive anchors that fit the sentence.

    line 72
  10. Distinctive:

    Read back the saved body and confirm the exact href, visible anchor, and page status.

    line 80
  11. Distinctive:

    Add relevant links to established coverage, including a useful hub when one exists.

    line 90
  12. Distinctive:

    Verify each target and keep the copy readable.

    line 91
  13. Distinctive:

    Keep older-page edits within the same published-source safety rules.

    line 93
  14. Distinctive:

    Use verified-seo-article-writing when the article also needs source and claim verification.

    line 95
  15. Distinctive:

    If a repeated insertion rule performs poorly or introduces irrelevant links, stop expansion and review the pattern before changing more …

    line 103
  16. Distinctive:

    Use distinct states: proposed, staged, published, verified, or blocked.

    line 114

“Common” is measured: kept by at least 6 in 10 of the 8 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Internal Linking

Build useful paths between pages, then prove that the intended links were saved. For a broad technical audit, start with seo-audit. For a recurring, lightweight opportunity harvest, use seo-quick-wins. Site hierarchy changes need a separate architecture plan.

Choose a mode

RequestMode
Find orphans, broken targets, or weak link coverageDiagnose
Choose targets, donors, and placementsPlan
Add approved links or rewrite anchorsImplement
Check results after a linking changeMeasure
Link a new article into the existing siteEditorial loop

Required inputs

  • Site and permitted URL scope.
  • Fresh crawl or page export with published status, body links, canonicals, and indexability where available.
  • Search or analytics data when ranking donors or measuring results.
  • Target page priorities and any editorial restrictions.
  • Permission scope: recommendations, draft edits, revisions, or approved live changes.

Missing data limits the result. Do not label a URL an orphan from a sitemap alone, or claim technical verification from a search snippet.

Guardrails

  1. New content stays Draft or Pending Review unless publication is authorized.
  2. Never change a published source to Draft to stage a link. Use a revision, staging copy, or authorized live edit.
  3. Count contextual body links separately from navigation, breadcrumbs, related-content modules, and footer links.
  4. Keep one planned body edge per source and target. Skip insertion if that body edge already exists, including a redirect or URL variant that resolves to the same canonical page.
  5. Use ordinary crawlable links. Do not add internal nofollow, sponsored, or ugc attributes without a documented reason.
  6. Verify the target returns a direct HTTP 200, is indexable, and has the intended canonical. Prefer linking directly to the canonical URL. If evidence is unavailable, mark the edge unverified and do not apply it.
  7. Quote a real placement from the current source. If it has disappeared, find a genuinely equivalent passage or reject the row. Never invent a claim to justify a link.
  8. Traffic alone does not make a good hub. Exclude off-topic, thin, duplicate, or unhelpful targets even when they receive visits.
  9. Start with a bounded manual batch. Bulk insertion requires explicit scope, rollback, and a small checked sample before expansion.
  10. Use evidence gathered after relevant canonical, redirect, or indexability fixes. Old exports may describe a problem that no longer exists.
  11. Read back the saved content and publication status. A successful save response is not proof that the link exists.

Diagnose

  1. Define the crawl scope, capture date, and coverage limits. Combine crawl discovery with published URL inventories to find pages the crawl cannot reach.
  2. Repair broken body targets and unnecessary redirect hops before adding more links.
  3. Identify pages with no discovered inbound body links. Separate true discovery problems from pages that should remain excluded or unindexed.
  4. Find relevant targets with search demand, reader value, or a conversion role. Use query and page data rather than position alone.
  5. Find potential donors with topical relevance, useful traffic, external links, or strong existing internal coverage.

Report counts for published URLs, unique body edges, orphan candidates, verified target issues, and coverage gaps. If a full crawl is unavailable, label the diagnosis partial.

Plan

Choose the applicable play after hygiene:

  • Authority to opportunity: relevant established donors link to useful pages that already show search visibility but need stronger internal discovery.
  • Traffic to conversion: visited pages link to relevant conversion pages. Use observed conversion data when available; do not invent conversion rates.
  • New to old and old to new: a new article links to established coverage, then older relevant pages are queued for links back to the new article.

Site search such as site:example.com "topic phrase" can discover candidates. It does not prove that a target is current, indexable, or relevant.

Rank edges by reader usefulness, target quality, strategic value, donor strength, and placement quality. Keep the first batch small enough to inspect every saved edit.

Return one row per unique edge:

SourceTargetPlayProposed anchorExact context quotePlacementEvidenceStatus

Use concise descriptive anchors that fit the sentence. Avoid empty anchors, "click here", forced exact matches, and repeated target links. Spread links where they help the reader. A link count is not a quota.

Implement

  1. Fetch the current source and target. Confirm the correct page identity and editing permissions.
  2. Recheck target status, indexability, canonical, existing body links, and the quoted placement.
  3. Stage the smallest necessary change. Preserve surrounding claims and formatting.
  4. Save through the permitted draft, revision, or live workflow.
  5. Read back the saved body and confirm the exact href, visible anchor, and page status.
  6. For staged changes to published pages, confirm the live page remains unchanged. For authorized live changes, inspect the rendered page and follow the link.
  7. Record rejected edges with the reason. Stop the batch if saves alter unrelated content or publication status.

Suggestion plugins may help find candidates. Verify current capabilities against current docs and treat suggestions as leads, not evidence. Keep automatic insertion disabled unless the approved workflow explicitly includes it.

Editorial loop

For each new article:

  • Add relevant links to established coverage, including a useful hub when one exists.
  • Verify each target and keep the copy readable. Do not pad with weak links.
  • Queue relevant older pages for inbound links after the new target is available and verified.
  • Keep older-page edits within the same published-source safety rules.

Use verified-seo-article-writing when the article also needs source and claim verification.

Measure

Capture a baseline and agree on follow-up windows before implementation. Compare touched pages with similar untouched pages where possible.

Record unique body inlinks, search impressions, clicks, query mix, position, relevant conversions, and implementation quality. Note seasonality, content changes, releases, and search-result changes. A before-and-after movement alone does not establish causality.

If a repeated insertion rule performs poorly or introduces irrelevant links, stop expansion and review the pattern before changing more pages.

Output contract

Return the edge queue plus:

  • Scope, evidence dates, and coverage limits.
  • Counts: proposed, rejected, staged, published, and verified edges.
  • Verification per applied edge: target checks, body href, anchor, page status, and readback result.
  • Remaining blockers and the next measurement date.

Use distinct states: proposed, staged, published, verified, or blocked. Report an unavailable check explicitly. Never report a staged revision as a live result.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.