docouture-documenting-changes
Skill246 starsChanged 46 days agoApache-2.0View on GitHub
How to update an existing docouture documentation site when a feature, change, deprecation or fix lands in the repo — figuring out what's affected from a diff/commit/PR, confirming with the user before writing, and keeping AGENTS.md's documentation-state ledger honest. USE WHEN a new feature was just added, something was deprecated or removed, a fix changes documented behaviour, or docs feel out of date relative to the code. EXAMPLES: 'I added a new CLI flag, update the docs', 'this API is now deprecated, reflect that in the docs', 'document this change', 'my docs are stale, what needs updating'.
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
If there's no site yet at all, this is the wrong skill — see docouture-getting-started first.
- Distinctive: line 17
Before considering ANY code change in this repo done — not just when the user explicitly asks to "update the docs" — check whether it needs …
- Very specific: line 27
No hit doesn't necessarily mean "skip it" either — the table is hand-maintained and can itself be stale (see reference/maintenance-loop.md).
- Distinctive: line 47
Find out what changed: Prefer inspecting the actual change first — a git diff, the commit(s) since the last documented pass, or a PR's …
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Documenting changes
This is the re-entry point for a docouture site that already exists (scaffolded and
drafted via docouture-getting-started) and needs to stay in sync as the repo evolves. Use
this every time something changes, not just once — it's the loop, not a one-shot pass.
If there's no site yet at all, this is the wrong skill — see docouture-getting-started
first.
Non-negotiable trigger check
Before considering ANY code change in this repo done — not just when the user explicitly asks to "update the docs" — check whether it needs this skill:
- List the file(s) you just touched.
- Grep the root
AGENTS.md's "Documentation state" table for each one, against every row'sderived fromcolumn (globs likeantora-extensions/lib/*.jscount — check the directory, not just the exact filename). - Any hit means that row's doc page now describes stale behaviour and this skill applies to this change, right now, as part of the same piece of work — not a follow-up, not something to wait to be asked for a second time.
- No hit doesn't necessarily mean "skip it" either — the table is hand-maintained and can
itself be stale (see
reference/maintenance-loop.md); a change to something clearly user-facing (a new CLI flag, a config key, a changed extension behaviour) still warrants a quick check of whether a doc page describes the old behaviour even if nothing in the table points at the file you touched.
This skill decides what changed and whether it needs a docs update; once it knows
which page(s) to touch, it hands off to the same skills docouture-getting-started does:
docouture-docs-internals— module/page/nav mechanics.docouture-writing-docs-pages— AsciiDoc authoring itself.docouture-authoring-guides— the touched page's content contract: what that section says, in what order, and its quality bar — so an edit to an existingreferenceorguidespage still matches its section's pattern rather than drifting from it.
See reference/maintenance-loop.md for the full mechanism: the AGENTS.md
documentation-state ledger, what counts as drift, and when to revisit the home page.
The flow, in short
- Find out what changed. Prefer inspecting the actual change first — a git diff, the commit(s) since the last documented pass, or a PR's description — over asking the user to enumerate it themselves. This is a starting point to propose from, not something to trust blindly: confirm your read of "what changed" with the user before writing anything, since inferred intent can be wrong (a refactor that touches many files is not the same as a new feature, even though both produce a large diff).
- Re-scan the repo's surface the same way
docouture-getting-started'scontent-sourcing.mdstep 3 describes — exports, CLI commands, config keys, API definitions — and diff it againstAGENTS.md's ledger. Seereference/maintenance-loop.mdfor exactly what counts asnew,staleandcurrent. - Check whether this needs a new page, an edit to an existing one, or a structural
change (a whole new module, if this is a monorepo gaining a new publishable artifact —
see
docouture-getting-started'sreference/structure-planning.mdfor that decision; this skill doesn't re-litigate structure on its own, it flags when structure planning is needed and hands back to that skill for the decision). - Write the update, following whatever page pattern already fits
(
docouture-docs-internals'sreference/page-patterns.md), the content contract for that page's section (docouture-authoring-guides), and the site's existing voice — a code-derived reference page reads differently from a hand-written guide, don't force one to imitate the other. - Update the ledger in
AGENTS.md— mark the touched page(s)currentagain, add a row for anything genuinely new, and never silently regenerate a page markedmanual(human-owned) without being asked. - Check
nav.adocfor the module(s) touched — a page that exists on disk but isn't listed there builds fine and is simply unreachable; nothing warns about this on its own.
What this skill deliberately doesn't do automatically
It never regenerates a whole site's structure from scratch on every change (that's
docouture-getting-started's job, and only ever needed once, or for a genuine restructure),
and it never writes past what a diff/commit actually supports without checking with the
user first — a hybrid of "look at the change" and "confirm before touching docs," not pure
guesswork and not pure Q&A.
In this repository56 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .github/copilot-instructions.mdCopilot instructions
- cavemanSkill
- design-an-interfaceSkill
- diagnoseSkill
- docouture-authoring-guidesSkill
- docouture-docs-internalsSkill
- docouture-docs-versioningSkill
- docouture-getting-startedSkill
- docouture-writing-docs-pagesSkill
- edit-articleSkill
- git-guardrails-claude-codeSkill
- grill-meSkill
- grill-with-docsSkill
- handoffSkill
- improve-codebase-architectureSkill
- migrate-to-shoehornSkill
- openspec-apply-changeSkill
- openspec-archive-changeSkill
- openspec-exploreSkill
- openspec-proposeSkill
- openspec-sync-specsSkill
- openspec-update-changeSkill
- openspec-verify-changeSkill
- prototypeSkill
- qaSkill
- request-refactor-planSkill
- reviewSkill
- scaffold-exercisesSkill
- setup-matt-pocock-skillsSkill
- setup-pre-commitSkill
- tddSkill
- teachSkill
- to-issuesSkill
- to-prdSkill
- triageSkill
- ubiquitous-languageSkill
- write-a-skillSkill
- writing-beatsSkill
- writing-fragmentsSkill
- writing-shapeSkill
- zoom-outSkill
- openspec-apply-changeSkill
- openspec-archive-changeSkill
- openspec-exploreSkill
- openspec-proposeSkill
- openspec-sync-specsSkill
- openspec-update-changeSkill
- openspec-verify-changeSkill
- openspec-apply-changeSkill
- openspec-archive-changeSkill
- openspec-exploreSkill
- openspec-proposeSkill
- openspec-sync-specsSkill
- openspec-update-changeSkill
- openspec-verify-changeSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
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.

