agentleFS
vstorm-covstorm-co/agenticos/.claude/skills/project-docs/SKILL.md

project-docs

Skill241 starsChanged 29 days agoApache-2.0View on GitHub

Write or fix the documentation site — add a page, update one after a behaviour change, fix a dead link or a failing docs build, add a diagram or an icon, or wire a new page into the nav. Use whenever docs/ or mkdocs.yml is touched, and whenever a code change alters behaviour a page describes. make docs-build runs --strict, so a dead link fails CI.

22 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use test-audit, to-issues and 3 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • BlockSkill

    test-audit

    Invoke whenever writing, changing, reviewing, or sweeping tests in buzz-app. Authoring gate for new tests plus audit workflow for low-value, implementation-coupled, or duplicative tests and the test-only production seams they demand.

    Also in 3 projects that use this

  • GoogleSkill

    to-issues

    Break a plan, spec, or PRD into independently-grabbable issues on the project issue tracker using tracer-bullet vertical slices.

    Also in 3 projects that use this

  • GoogleSkill

    to-prd

    Turn the current conversation into a PRD and publish it to the project issue tracker — no interview, just synthesis of what you've already discussed.

    Also in 3 projects that use this

  • Google CloudSkill

    triage

    Gather context on a change or bug and plan the implementation.

    Also in 3 projects that use this

  • BlockSkill

    frontend-design

    Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 4 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 18 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Write it in docs/ (or docs/howto/, docs/reference/).

    line 117
  2. Very specific:

    Add it to nav in mkdocs.yml, at the position it belongs at — a page outside the nav is a build warning under --strict.

    line 118
  3. Distinctive:

    Use second person and present tense for user instructions.

    line 125
  4. Distinctive:

    Use sentence case headings, preserving product names and acronyms.

    line 129
  5. Very specific:

    Keep paragraphs within the limit enforced by scripts/check_docs_paragraphs.py.

    line 130
  6. Distinctive:

    Add a recap only when it helps a reader retain the decisions from a long page.

    line 132
  7. Distinctive:

    Do not repeat short pages or force a fixed number of bullets.

    line 133

“Common” is measured: kept by at least 6 in 10 of the 18 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

The docs site

make docs                       # serve on :8001 (DOCS_PORT=8002 if it is taken)
make docs-build                 # --strict; CI runs this, a dead link fails the build

docs/ is both the published site and the repository's own engineering notes — one copy, on purpose. A second copy of "how a capability works" written for outsiders is a copy that disagrees with the one contributors read. So: do not write a page that restates another page. mkdocs.yml's nav decides what an outside reader meets first; nothing is duplicated for them.

The same principle governs these skills. A skill routes to the doc and adds the operational layer; it does not paraphrase the doc.

Where to write when behaviour changes

The table in AGENTS.md under ## Documentation maps topic → page. Keep it current when you add a page.

Three things that silently render wrong

These have all shipped broken. Check the rendered page, not the Markdown.

Mermaid needs the custom fence. Without custom_fences on pymdownx.superfences, a ```mermaid block goes to the syntax highlighter and the reader gets the graph's source in a code box. Configured now — do not remove it.

Verifying is harder than it looks: Material renders into a closed shadow root, so the div.mermaid in the DOM is empty and querySelector('.mermaid svg') finds nothing. That is indistinguishable from a broken render. A screenshot is the only honest check. Mermaid itself loads from unpkg.com at page-view time, so diagrams do not render offline or behind a strict CSP.

Icons need pymdownx.emoji. Without the twemoji index, :material-download:{ .lg .middle } reaches the page as that literal string — the home page's grid cards rendered their own source for a while.

--strict does not validate anchors. A #fragment that matches no heading passes the build. Check fragments by hand, or with a script over the headings.

Search and link previews

scripts/mkdocs_hooks.py writes what a search result or a shared link reads from a page's head; Material writes only the title, description, canonical and hreflang.

  • seo_title in front matter replaces the <title> - at most 60 characters, starting with the page's own name. title stays short because the nav uses it.
  • description is the meta description and the card text - at most 160 characters. A page without one falls back to site_description.
  • Every page gets Open Graph and Twitter tags, with docs/assets/social-preview.png as the image from the site root in every locale.
  • A ## Frequently asked questions section becomes FAQPage structured data: each ### inside it is a question, the text up to the next one its answer. A translation must pin { #frequently-asked-questions }, which is how the hook finds the section in every language. Answers are published as plain text, so keep each one self-contained. backend/tests/test_mkdocs_seo.py covers the extraction.

Translate seo_title and description with the page; they are read by people searching in that language.

mkdocstrings

The API reference is generated from docstrings rather than written twice — this codebase puts its reasoning in docstrings, so a hand-written reference would be a worse copy of something already there.

The collector is static, and app/services/, app/api/ and app/worker/ have no __init__.py, so it cannot traverse into them: ::: app.services.foo fails the build. Reference those from prose with a source link until those packages are made explicit.

Because the reference is generated: when behaviour changes, fix the docstring. A prose page that repeats it will drift.

The seven tabs

AgenticOS · Features · Learn · Reference · Resources · About · Release Notes — FastAPI's arrangement, organised by what the reader is doing rather than by what a page is.

Learn is a sequence, not a bag: Get started → Build the agent → Put it in front of people → Keep it under control → the recipes. A page added there belongs at a position.

Files stay flat in docs/. The nav does the grouping. Those paths are named in AGENTS.md, in these skills, in scripts/docs_drift.py and in backend docstrings, so moving a file to get a prettier URL invalidates all of them.

release-notes.md is CHANGELOG.md, substituted at build time by scripts/mkdocs_hooks.py. That is a hook rather than a snippets include because snippets expand after every hook, so nothing could reach the included text to rewrite its docs/-prefixed links — and --strict fails on the docs/docs/… they resolve to.

Two pages are in the repository and not on the site

docs/ROADMAP.md and docs/about/design.md sit in exclude_docs beside design/, plans/ and audits/. Delivery state written in board shorthand and the six decisions a contributor needs before their first change are read by somebody who has already chosen to work on this; published, they put an internal status report in a tab a prospective user meets first.

A published page that links to either one fails --strict, because an excluded file is not built and the link resolves to nothing. Link them by GitHub blob URL instead - docs/about/index.md and docs/reference/capabilities.md are the two that do.

Adding a page

  1. Write it in docs/ (or docs/howto/, docs/reference/).
  2. Add it to nav in mkdocs.yml, at the position it belongs at — a page outside the nav is a build warning under --strict.
  3. Cross-link it from the pages a reader arrives from, and from AGENTS.md's table.
  4. make docs-build and check the anchors.

Voice

Use second person and present tense for user instructions. Explain behaviour and relevant reasons; distinguish verified facts from uncertainty. Choose prose, lists or tables according to what helps the reader.

  • Use sentence case headings, preserving product names and acronyms.
  • Keep paragraphs within the limit enforced by scripts/check_docs_paragraphs.py. Split by idea; use an admonition when a reader needs a warning before acting.
  • Add a recap only when it helps a reader retain the decisions from a long page. Do not repeat short pages or force a fixed number of bullets.

Icons

app/core/catalog/icons/<name>.svg is served by GET /catalog/icons and drawn for a catalog entry or provider whose id matches and which no compiled-in icon set carries. The file's own colours are ignored — it renders as a currentColor silhouette, unlike the compiled-in marks, which carry their brand's colour where it reads on both themes. See icons/README.md.

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.