handsofflinks / what developer registries serve
First-party measurement, read 15 Aug 2026
What rel attribute developer registries actually serve
Five package registries, read off the served HTML rather than the docs. The attribute changes depending on where the link sits on the page, and most write-ups miss that.
Short answer up top. On the package page for github.com/gorilla/mux, the
structured Repository link is served with noopener and carries no
nofollow, sponsored or ugc value. Links rendered from the README body on that same page come
back with nofollow
(pkg.go.dev package page, read 15 Aug 2026).
One page, two attributes. What changes is where the link sits.
Google says links marked with the sponsored, ugc or nofollow rel attributes will generally not be followed, and defines nofollow as the value to use when you would rather Google not associate your site with, or crawl, the linked page (Google Search Central). Nothing here promises a ranking or traffic outcome. These are readings, not predictions.
What each registry served
Every row was read off the served HTML of the live page on 15 August 2026.
| Registry | Which link | rel served | Read from |
|---|---|---|---|
| pkg.go.dev | structured Repository link | noopener | pkg.go.dev/github.com/gorilla/mux |
| pkg.go.dev | link inside the rendered README body | nofollow | pkg.go.dev/github.com/gorilla/mux |
| docs.rs | crate sidebar homepage link | no rel attribute | docs.rs/serde |
| docs.rs | external link in the documentation body | no rel attribute | docs.rs/serde |
| JSR | structured GitHub repository chip | noopener noreferrer | jsr.io/@std/path |
| JSR | link rendered from the README body | nofollow | jsr.io/@std/path |
| code.dlang.org | package Homepage link | no rel attribute | code.dlang.org/packages/vibe-d |
| code.dlang.org | external link from the package README | no rel attribute | code.dlang.org/packages/vibe-d |
| github.com profile | the website field on a user profile | nofollow me | github.com/sindresorhus |
Across the registry pages probed for this page, read as Googlebot, the share of structured repository or homepage links served without a nofollow, sponsored or ugc value came out at 100 percent. That rate covers the rows in the table above and nothing wider. README-body links are excluded from it, and two of the four registries nofollow theirs.
Read any anchor you fetched
Paste the raw anchor element off a page you fetched yourself. The reader reports the rel tokens it finds and whether any is one of the three values Google names. It runs in your browser and sends nothing anywhere.
A followed link still needs a crawl path
Google uses links both as a relevancy signal and as the mechanism for finding new pages to crawl, and says it can generally only crawl a link if it is an anchor element with an href attribute (Google Search Central). So the attribute is half the question. The other half is whether anything reaches the page carrying it.
pkg.go.dev declares an XML sitemap index in its robots.txt and disallows only
/search?* and /fetch/* for all user agents, which leaves package
documentation pages crawlable
(pkg.go.dev/robots.txt).
jsr.io also declares a sitemap, and its Content-Signal line says search indexing is permitted while AI training is not (jsr.io/robots.txt).
code.dlang.org applies Disallow: / to a long list of AI and data-collection
crawlers including GPTBot, ClaudeBot, CCBot, Bytespider and GoogleOther. Googlebot itself is
not listed (code.dlang.org/robots.txt).
Now the contrast that matters. A published Telegraph article is server rendered, with the
article title present in the raw HTML without JavaScript execution, and it carries a
permissive index, follow robots meta tag
(telegra.ph sample article).
But the served HTML of that article contains no navigation, index or related-post links to other telegra.ph URLs. The only anchor on the page is a mailto DMCA link, so the page offers no internal path to any other Telegraph page.
telegra.ph publishes no robots.txt and no sitemap either. Both telegra.ph/robots.txt and telegra.ph/sitemap.xml returned HTTP 404 on 15 August 2026 and served HTML rather than a robots file or XML.
njump.me is the other shape of the same problem. Its robots.txt blocks DataForSeoBot,
SemrushBot, MJ12bot, dotbot, PetalBot and Amazonbot with Disallow: /, so links
published there won't appear in backlink indexes built by those crawlers. A closing wildcard
rule permits all other user agents
(njump.me/robots.txt).
It publishes no sitemap. njump.me/sitemap.xml returned HTTP 404 and an HTML error document on 15 August 2026.
Probes that failed, printed at the same size
crates.io returned HTTP 403 to the scripted request this page's probe made against its
api/v1/crates endpoint and cited its API data access policy in the error body
(crates.io/api/v1/crates/serde), so no
crates.io row appears in the table above. Re-reading it on 15 August 2026 returned HTTP 200 to
several other user agents, so treat that block as agent-specific rather than a property of
scripted access in general.
No live-of-built ledger count for any of these surfaces is present in this page's validated fields, so none is printed here rather than estimated.
Where these readings come from
Attributes were read off the served HTML of each live page, not from documentation describing those pages.
The pkg.go.dev fetch was repeated with the Googlebot user agent on 15 August 2026 and
returned the identical Repository anchor, including rel="noopener", that a desktop
browser user agent received. So the attribute isn't varied by user agent on that page
(pkg.go.dev/github.com/gorilla/mux).
Rates on this page are recomputed in code from the fetched files, never typed by hand. Any figure without a file or a cited page behind it is left out.
Common questions
Does a missing rel attribute mean the link is followed?
Google names sponsored, ugc and nofollow as the values whose links will generally not be followed (Google Search Central). What the probe observed on docs.rs and code.dlang.org is the absence of any rel attribute on the sidebar homepage link and the package Homepage link. That's a reading of the served markup, not a statement about what Google did with it.
Why does a README link differ from the repository link on the same registry?
Both pkg.go.dev and JSR serve their README-body links with rel="nofollow"
while the structured repository link on the same page carries noopener or noopener
noreferrer. The observed difference tracks where the link sits in the template. Neither page
states a reason, and none is inferred here.
What does Google call link spam?
Google defines link spam as creating links to or from a site primarily for the purpose of manipulating search rankings, and lists using automated programs or services to create links to your site as an example, along with exchanging money for links or for posts that contain links (Google Search Central spam policies). Read the policy yourself and decide.
Is a followed link useful on a page nothing links to?
Google uses links to find new pages to crawl. A Telegraph article carries an index, follow robots meta tag, yet its served HTML contains no links to any other telegra.ph URL apart from a mailto DMCA link. Those two readings sit next to each other on the same page, and the second one is why a rel attribute alone is not a channel.
What handsofflinks is
A self-hosted pipeline that builds linkable assets from your own numbers and publishes them through public platform APIs under accounts you own. It re-fetches each published page, reads the rel attribute off the served HTML and confirms a crawl path before any ledger row may say live. It's macOS only and runs in a terminal, and hands-off autonomous operation needs an agentic CLI. There's no outreach email in the system and no ranking is promised.
Disclosures
No ranking or traffic outcome is promised on any page. Ledger figures are the operator's own history with this method across 16 of his own sites over 37 days, and they aren't a forecast of any buyer's results. The shipped build v1.0.5 is a rewrite of the code that produced that history and has 2 independently verified live links of its own.
Search Console showing no manual action is the operator's account of the operator's own sites, which a reader cannot audit and which is not evidence about any other site. Competitor prices are dated snapshots read from each seller's own public page on the date shown, and the average guest post figure comes from a marketplace that sells guest posts.
Domain Rating and linking-website counts for belikenative.com, zovo.one and ingredientcalculator.com are whole link profiles built over years by every channel their owner has used, not output of this pipeline. Hands-off autonomous operation requires an agentic CLI such as Claude Code, which is a recurring cost paid to that vendor and not to handsofflinks.