Skip to content

docs(skills): add docsite-link-sweep maintainer skill - #4071

Draft
twishabansal wants to merge 2 commits into
mainfrom
docs/add-docsite-link-sweep-skill
Draft

twishabansal wants to merge 2 commits into
mainfrom
docs/add-docsite-link-sweep-skill

Conversation

@twishabansal

@twishabansal twishabansal commented Sep 17, 2026

Copy link
Copy Markdown
Contributor

Description

Adds a docsite-link-sweep maintainer skill: it sweeps the docs for broken and non-canonical links, reports each finding with the reason it breaks, and applies the safe class of internal link fixes. It exists because the repo's two checkers disagree. lychee resolves links as filesystem paths and knows nothing about Hugo; Hugo resolves .md links to pretty URLs and generates whole classes of links from shortcodes, but never checks an external URL. The skill targets that gap: directory-style links lychee cannot resolve, site-absolute links that leak out of /dev/ and versioned builds, shortcode-generated links that exist only in built HTML, and links into ignoreFiles paths that resolve on disk but 404 live. Fixes are limited to five mechanical rewrite classes, each verified against both checkers; everything else is reported, not applied. Tracked in toolbox-skills/ideas.md.

PR Checklist

  • Make sure to open an issue as a bug/issue before writing your code!
  • Ensure you have manually reviewed the entire diff before requesting a review
  • Ensure the tests and linter pass
  • Code coverage does not decrease (if any source code was changed)
  • Appropriate docs were updated (if necessary)
  • Make sure to add ! if this involves a breaking change

Sweeps the docs for broken and non-canonical links, reports each finding
with the reason it breaks, and applies the safe class of internal link
fixes to the working tree.

The skill exists because the two checkers guarding these docs disagree:
lychee resolves links as filesystem paths and knows nothing about Hugo,
while Hugo resolves .md links to pretty URLs and generates whole classes
of links from shortcodes but never checks an external URL. A link can
pass one and break the other.

Covers four failure modes neither checker reports on its own: directory
style links that lychee cannot resolve, site-absolute links that leak out
of /dev/ and versioned builds to the latest-release docs, shortcode
generated links that appear only in built HTML, and links into ignoreFiles
paths that resolve on disk but 404 on the site.

Every command in the skill was run against this repo before being written
down.

@gemini-code-assist gemini-code-assist Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request introduces a new AI agent skill, docsite-link-sweep, along with supporting reference documentation (link-forms.md and a reference to DEVELOPER.md) to help maintainers sweep the mcp-toolbox documentation for broken and non-canonical links. The review feedback suggests replacing absolute GitHub URLs pointing to repository configuration files (such as .lycheeignore, workflow files, and hugo.toml) with workspace-relative paths, ensuring the local AI agent can read the current state of the workspace files directly without making unnecessary network requests.

Comment thread skills/maintainer/docsite-link-sweep/SKILL.md Outdated
Comment thread skills/maintainer/docsite-link-sweep/SKILL.md Outdated
…ychee

Every lychee workflow in the repo is currently disabled_manually, so the skill no longer claims CI checks links. Step 1 now queries workflow state and tells the sweep to report when it is the only link checking happening.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant