Sync site docs
Codex non-officiel sur Claude Code — theorie, ecosysteme, case studies, ressources. Code en pair-programming avec Claude Code.
npx -y skills add Sylad/claude-code-codex --skill sync-site-docsAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Use after a notable change on one of the 6 personal sites (new nav entry, new page, architecture pivot, removed feature, new third-party tool tested) to keep the project docs and the cross-site case study in sync. Triggers on "/sync-site-docs", "sync docs", "j'ai changé la nav", "j'ai ajouté une feature importante", "update doc projet", or after a substantial commit on warhammer40k / finance-tracker / ol-companion / avatar-pandora / evatosorus / claude-code-codex.
SKILL.md
7.9 KB, ~1.9k tokens by cl100k_base, as published. Nobody here has run it
Sync Site Docs
Keeps the documentation of the 6 personal sites coherent across two layers :
- Project-local docs —
CLAUDE.md,README.md, and any pertinent memory note (<app>_state_*.md,<app>_imagery.md, etc.) inside the touched repo. - Cross-site case study — the matching page on
claude-code-codex/frontend/src/pages/case-studies/<slug>.astro(architecture Mermaid diagram, Pivots & lessons learned section, Tools tested if a new third-party tool was just experimented with).
When to use
- ✅ The site got a new top-level page (e.g.
/startwas added on claude-code-codex — README pages table needs an update, the case-study page should mention it). - ✅ A menu/nav was reorganized.
- ✅ An architecture pivot happened (new backend, new data source, removed feature, swapped framework).
- ✅ A new third-party tool got tested in sandbox (Ruflo-style) —
the case-study
/case-studies/ruflopattern lives on, plus a Watchlist memo update. - ✅ A new live feature went into prod that visitors will see.
When NOT to use
- ❌ Tiny fixes (typos, copy tweaks, color shift) — no doc impact.
- ❌ Refactors that don't change the user-visible surface — internal cleanups don't need cross-doc sync.
- ❌ The change hasn't been committed yet — work on a clean tree, this
skill reads
git logto detect what landed.
Procedure
Track each step with TodoWrite from the start.
Step 1 — Identify the site
git rev-parse --show-toplevel
Map repo root to one of the 6 known sites :
| Repo root | Site |
|---|---|
~/projects/developpeur/warhammer40k | warhammer40k |
~/projects/developpeur/finance-tracker | finance-tracker |
~/projects/developpeur/ol-companion | ol-companion |
~/projects/developpeur/avatar-pandora | avatar-pandora |
~/projects/developpeur/evatosorus | evatosorus |
~/projects/developpeur/claude-code-codex | claude-code-codex |
If the cwd isn't a known site, ask the user to cd to the right repo.
Step 2 — Detect what changed
git log --oneline -15
git diff HEAD~5..HEAD --stat
If the user passed an argument-hint, use it as the lead. Otherwise infer from the diff. Typical signals :
frontend/src/pages/<new>.astrocreated → new top-level pageNavbar.vuemodified → nav restructuringpackage.jsondeps added/removed → stack changefrontend/src/data/<new>.ts→ new data sourcescripts/sources/<new>.ts→ new batch source- File deletions → feature removed
If nothing significant landed, abort with a friendly note.
Step 3 — Audit the project-local docs (decide if update needed)
For the touched site, read :
CLAUDE.md(project root) — does any rule need adding/updating ?README.md— is the pages table / stack section current ?- Memory files at
~/.claude/projects/<project>/memory/<app>_*.md— especially<app>_state_*.mdfiles which capture the current snapshot.
For each file, decide :
- No change — content still accurate.
- Append — new info to add (new page, new tool).
- Replace section — old info now wrong (architecture, removed feature).
Step 4 — Audit the cross-site case study
If the touched site has a case study at
~/projects/developpeur/claude-code-codex/frontend/src/pages/case-studies/<slug>.astro :
Read that file. For each change identified in Step 2, decide :
- Architecture diagram (Mermaid) — outdated nodes/arrows ?
- Stack & data flow — outdated tech mention ?
- Process avec Claude Code — new pattern used worth mentioning ?
- Pivots & lessons learned — recent abandonment / new direction worth capturing ?
Special cases :
- claude-code-codex itself has its case study at
case-studies/claude-code-codex.astro(meta). Don't forget it. - A third-party tool tested (Ruflo-style) lives at
case-studies/<tool>.astro. The Tools tested section index oncase-studies.astroand the entries infrontend/src/data/tools-tested.tsmay also need an update.
Step 5 — Confirm with the user
Present the diff of what you'd write via AskUserQuestion. Each
proposed change as one option (or grouped if they're tightly coupled).
Let the user :
- ✅ Approve all
- ✏️ Edit specific items (re-word, drop)
- ❌ Reject specific items
Don't write anything yet.
Step 6 — Apply the approved changes
For each approved diff :
Editfor in-file replacements (preserving the rest)Writeonly when creating a new memory note- For the case study page, prefer surgical
Editover rewrite — the 6 sections (Pourquoi / Architecture / Stack / Process / Pivots / links) keep their structure.
Step 7 — Cross-repo coordination
If the change affects both the touched site repo AND claude-code-codex, you may end up with edits in two repos. Don't auto-commit on the user's behalf — at the end, present :
Modified in <site-repo> :
- CLAUDE.md (+5 lines)
- README.md (pages table updated)
Modified in claude-code-codex :
- frontend/src/pages/case-studies/<slug>.astro (Pivots section + diagram)
- frontend/src/data/tools-tested.ts (Ruflo entry tweaked)
Suggested commits :
cd <site-repo> && git add … && git commit -m "docs: …"
cd ~/projects/developpeur/claude-code-codex && git add … && git commit -m "docs(case-study): …"
The user runs the commits and pushes when ready.
Examples
Example A — claude-code-codex got a new /start page
- Project docs:
README.mdpages table gains the new row + skills section. - Case study (
/case-studies/claude-code-codex.astro) — the "Sections livrées" section in Pivots can mention the new starter page; the Mermaid diagram stays unchanged (no archi shift). - Memory :
claude_code_codex_kickoff.mdalready lists pages — append.
Example B — warhammer40k removes the Sectors page
- Project docs :
CLAUDE.mdalready lists "Sectors abandonné" (cfwarhammer_sectors_abandoned.md). No change needed there. - Case study (
/case-studies/warhammer40k.astro) — remove Sectors from the architecture diagram + add a one-liner in Pivots.
Example C — finance-tracker adds a new bank parser
- Project docs :
CLAUDE.mdmay gain a rule on the new file naming convention.README.mdif the visible scope changes. - Case study : Stack & data flow gains the new parser; if the heuristic is non-obvious, add a Pivots line.
Anti-patterns
- ❌ Updating doc files without checking git history first — you might re-add things that were intentionally removed.
- ❌ Forgetting one of the two layers (project-local OR cross-site case study). The skill exists precisely to cover both.
- ❌ Auto-committing without showing the user what changed first.
- ❌ Spamming
AskUserQuestionwith 10 micro-edits — group related changes into 2-4 questions max. - ❌ Touching the global
~/.claude/CLAUDE.md— that's user-level, outside this skill's scope. Project-local CLAUDE.md only.
Companion skills
/save— captures meta learnings of a session ; this skill captures concrete site-level changes. Often run together./skillify— when the change you want to document is a tool/concept that should also become a skill of its own.
Gives 0 of the 12 instructions most docs writing skills give in ~1.9k tokens
Counted across 1,637 of the 3,044 authors here whose files we hold, read 2026-08-06
- announce the skill at startin 54 of 1637, across 21 files
- convert legacy doc files before editingin 45 of 1637, across 7 files
- predict questions readers might askin 42 of 1637, across 3 files
- Generate clarifying questions for initial contextin 42 of 1637, across 3 files
- Create document scaffold with placeholder textin 42 of 1637, across 3 files
- Brainstorm content options for each sectionin 42 of 1637, across 3 files
- Test document with fresh context-less instancein 42 of 1637, across 3 files
- ask interview questions one at a timein 42 of 1637, across 26 files
- include exact file paths in every taskin 42 of 1637, across 15 files
- Apply surgical edits during refinementin 41 of 1637, across 2 files
- Offer structured workflow or freeformin 40 of 1637, across 1 file
- Ask for document meta-contextin 40 of 1637, across 1 file
Said here and by no other author read
- track each step with TodoWrite
- identify the touched site from the repo root
- read project-local documentation files
- read the cross-site case study file
- propose changes to the user before writing
- group related proposed changes into few questions
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.