agentleFS
gnegreiros-uxgnegreiros-ux/agentica-design-system/.agents/skills/document/SKILL.md

document

Skill2 starsChanged 45 days agoApache-2.0View on GitHub

Post-work documentation orchestrator for the Agentica repository. Given what was just done in the session, checks every documentation surface (guidelines, ADRs, git-workflow rules, site changelog, GitHub Projects, npm/starter-kit READMEs, component stories), writes the relevant ones, verifies the build/tests, and proposes — never executes — a commit and push. Use after a chunk of work is functionally done, to catch up documentation surfaces that are easy to forget.

  • Commits and pushes

56 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use develop, scope and 10 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.
  • hong9LeeSkill

    develop

    빌드, 아키텍처, git 관련 작업 시 OMS MSA 공통 인프라 규칙을 적용한다. 기술 스택, 빌드 명령, 헥사고날 아키텍처 구조, git 워크플로우, AI Context 동기화 규칙을 다룬다. 코드 작성 컨벤션은 /convention, 테스트 컨벤션은 /test-guide 스킬을 참조.

    Also in 5 projects that use this

  • YuriNachosSkill

    scope

    Configure claude-scope status line widgets for Claude Code. Use when user wants to customize widgets, change theme, reorder elements, or modify display style.

    Also in 5 projects that use this

  • ralphkraussSkill

    resolve-conflicts

    Resolve git merge conflicts or re-apply changes from a stale branch onto current main. Use when user says "resolve conflicts", "merge conflicts", "rebase branch", "branch is behind", "re-apply changes", "port branch", or a branch is too far behind main to merge cleanly.

    Also in 3 projects that use this

  • MicrosoftSkill

    sync

    Sync the current session branch with its upstream branch, or publish the current session branch to a remote. Use when the user asks to sync a branch, pull latest changes, rebase onto upstream, push current branch, publish branch, or set upstream.

    Also in 5 projects that use this

  • kt057Skill

    check

    lint, typecheck, test を一括実行して結果をサマリーする

    Also in 7 projects that use this

  • Microsoft AzureSkill

    test

    Also in 12 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 52 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. Distinctive:

    Do not document something that didn't actually happen, and do not infer scope beyond what the diff and conversation actually support.

    line 55
  2. Very specific:

    Read the full trigger matrix in .claude/skills/pipelines/docs.md — the canonical, single-source-of-truth list of every documentation …

    line 60
  3. Very specific:

    Follow that surface's own established conventions (tone, structure, bilingual FR/EN spans on the site, Conventional Commits phrasing in …

    line 83
  4. Distinctive:

    Prefer updating an existing entry (a ticket, a changelog section) over creating a duplicate when the work is a continuation of something …

    line 86
  5. Distinctive:

    Keep entries factual and specific (file paths, PR/ADR numbers, concrete outcomes) over vague summaries — matches this repository's existing …

    line 88
  6. Distinctive:

    Run whatever subset of these actually applies to what was touched in Step 3:

    line 93
  7. Very specific:

    If branch protection currently forbids direct pushes to the target branch (enforce_admins: true, see ADR-077), the proposed action is a …

    line 114

“Common” is measured: kept by at least 6 in 10 of the 52 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

Skill: document

Cross-tool port of .claude/skills/document/SKILL.md (Claude Code's native implementation, invoked there as /document) — see governance/ai-skills-reference.md for the full picture of what Agentica's skills automate. This file follows the open Agent Skills format (agentskills.io) so Codex CLI and GitHub Copilot CLI load it automatically from .agents/skills/ when the task looks like "document what we did" / "catch up the docs" / "what still needs updating" — there's no /document slash command outside Claude Code, so rely on the description above for automatic matching, or invoke it explicitly if your tool supports that.

Absolute rule

No commit or push happens automatically. This skill always ends in a proposal, never in an executed commit. Same governance boundary as quality-gate and governance/rules/git-workflow.md — the human always has the final word.

Relationship to quality-gate

quality-gate is the full pre-commit gate: token consistency, WCAG, UX pattern review, ADR compliance, missing-ADR triggers, documentation, site rebuild, commit — run for every change, however small. This skill is narrower and lighter: it only walks the documentation surfaces (via .claude/skills/pipelines/docs.md — plain Markdown, readable regardless of which tool you're using) and verifies the build/tests don't break. Reach for this after a chunk of work is functionally done and already committed-worthy, specifically to catch up documentation surfaces that are easy to forget because they live outside the file(s) actually changed — GitHub Projects and the site changelog above all. It does not replace quality-gate for changes that also touch tokens, components, or site pages directly — run both when in doubt.

Trigger

Invoked when asked to document recent work, optionally with a short description of what to document (e.g. "document the npm publication chantier"). With no specific scope given, the scope is "everything done in this session that isn't reflected in documentation yet" — reconstructed from conversation context and git log/git diff against the last pushed commit on the current branch.

Process

Step 1 — Establish what was actually done

Build a factual list of the session's changes before touching any documentation:

  • git log --oneline since the branch diverged from main (or since the last push)
  • git diff --stat for anything still uncommitted
  • Conversation context for intent/rationale a diff alone won't show (why a decision was made, what was explicitly rejected, what the user asked for)

Do not document something that didn't actually happen, and do not infer scope beyond what the diff and conversation actually support.

Step 2 — Check every place against the docs pipeline

Read the full trigger matrix in .claude/skills/pipelines/docs.md — the canonical, single-source-of-truth list of every documentation surface in this repository (a plain Markdown file, readable regardless of which AI tool you're using — it isn't Claude-specific content, just historically filed under .claude/). For each row, decide relevant/not relevant against Step 1's list, and say so explicitly (a place that's genuinely not applicable is reported as "N/A", never silently skipped).

Two surfaces are the most commonly missed because they live outside the changed files entirely — always check them explicitly, even for small changes:

  • GitHub Projects (Backlog/board) — was a chantier completed, started, or newly identified?
  • Site changelog (buildChangelog() in site/build.js) — was anything user-visible shipped (feature, fix, decision)?

Also check .claude/skills/pipelines/adr-triggers.md explicitly — governance changes (new CI/CD pipeline, new governance rule) require a new ADR even when no token, component, or site file changed. An ADR that reverses or amends a still-Active ADR is a new ADR referencing the one it amends — ADRs are immutable once active, never edited in place (see decisions/README.md's registry rules).

Step 3 — Document

For each relevant surface identified in Step 2:

  • Follow that surface's own established conventions (tone, structure, bilingual FR/EN spans on the site, Conventional Commits phrasing in changelogs, the Status/Domaine/Date/ADR fields on GitHub Projects per ADR-069)
  • Prefer updating an existing entry (a ticket, a changelog section) over creating a duplicate when the work is a continuation of something already tracked
  • Keep entries factual and specific (file paths, PR/ADR numbers, concrete outcomes) over vague summaries — matches this repository's existing documentation style

Step 4 — Verify nothing is broken

Run whatever subset of these actually applies to what was touched in Step 3:

  • node site/build.js — required if any site page or the changelog changed; confirm 0 orphaned CSS variables and no build error
  • node scripts/audit-language.js --ci — required if any bilingual site content was added or a non-site file with English-only requirements was touched. Run it after git add on any new file, not before: the script scans tracked files only, so a new untracked file passes a pre-add local run clean and only fails once CI sees it as tracked
  • Structural HTML validation on any page whose template changed (balanced tags, no orphaned wrapper divs)
  • Bilingual FR/EN parity check (lang-fr / lang-en span counts) on any site page edited
  • A GitHub Projects item, once created or edited, is read back (gh project item-list/item-view) to confirm the field values actually saved — the API has silently rejected malformed field updates before

Step 5 — Report and propose

Present a report in the format below, then stop and wait for explicit approval before staging, committing, or pushing anything — exactly like quality-gate.

If branch protection currently forbids direct pushes to the target branch (enforce_admins: true, see ADR-077), the proposed action is a feature//docs/ branch + PR + merge, not a direct push — say so explicitly in the proposal rather than defaulting to git push.

Report format

## Documentation report

### What this covers
[1-2 sentences: the chunk of work being documented, with commit range or PR references]

### Documentation surfaces checked
- [x] Site changelog — added 3 bullets under "npm publication" (Unreleased)
- [x] GitHub Projects — ticket #NNN created/updated, status → Terminé <!-- lang-audit-ignore: literal GitHub Projects Status field value, which is French in this project -->
- [ ] guidelines/components/*.md — N/A, no component behavior changed
- [ ] decisions/ADR-0XX.md — N/A, no new architectural decision
- [x] packages/tokens/README.md — N/A, already up to date
- ...(every row of pipelines/docs.md gets a line, including the N/A ones)

### Verification
- [x] node site/build.js — 109 files generated, 0 orphaned CSS variable
- [x] Bilingual parity — 132/132
- [x] GitHub Projects fields confirmed via item-list read-back

### Proposed commit
[branch name, commit message following governance/rules/git-workflow.md conventions]

Proceed with commit and push?

What this skill does NOT do

❌ Commit or push without explicit approval — reports and proposes only
❌ Invent facts, PR numbers, or outcomes not actually verifiable from the diff/conversation
❌ Skip a row in pipelines/docs.md without stating why it's N/A
❌ Create a duplicate GitHub Projects ticket for work already tracked elsewhere
❌ Run the full quality-gate pipelines (tokens/WCAG/UX/ADR compliance) — use
   quality-gate directly when those are actually in scope

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.