wiki
Skill2 starsChanged 5 days agoMITView on GitHub
Create and maintain an organic project wiki — updates pages affected by recent work, scans for gaps and staleness, reconciles the index. Shape emerges from the project, not a template.
- Reads credentials
- Deletes or force-pushes
- Installs packages
45 other projects use a version of this.
Adopted by
Most recently taken up by
- cz-chan/skillstheir own version, first seen Nov 16, 2026
- Yomudogly/oh-my-agytheir own version, first seen Nov 16, 2026
- tboome33/obsidian-mcp-routertheir own version, first seen Nov 15, 2026
- Yeachan-Heo/oh-my-claudecodetheir own version, first seen Nov 9, 2026
- skillberry-ai/cap-evolvetheir own version, first seen Nov 8, 2026
- WALKERKILLER/Look-Tongji-Notestheir own version, first seen Nov 8, 2026
Used with
Projects with it also use loft, wiki-fold 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.
Yeachan-HeoSkillloft
Loft the shape before cutting steel — answer a design question that prose cannot settle by building a throwaway artifact: a pure logic module in a clickable shell, or structurally different UI variants behind one route. The captain reacts to the artifact; the answer folds into the decision; the artifact never docks. Use when a design question stalls in words, when a navigator map carries a loft ticket, or when a spec discussion reaches "we would have to see it".
Also in 3 projects that use this
tboome33Skillwiki-fold
Roll up the wiki's journal.md entries into structured fold pages — like 2^k log compaction. Reads the last 2, 4, 8, 16... entries and writes a fold page that summarizes them by extractive summarization (no invention), with backlinks to children. Idempotent at the structural level — re-running with the same window produces a byte-equivalent fold. Use when the user says "fold the log", "rollup recent activity", "compact the log", "/wiki-fold", or after many ingestions when journal.md is becoming too long to read directly.
Also in 3 projects that use this
Jurel89Skillai-slop-cleaner
Clean AI-generated code slop with a regression-safe, deletion-first workflow and optional reviewer-only mode
Also in 5 projects that use this
ibahgatSkillralplan
Create a strict, quality-first execution plan with explicit gates and rollback-ready sequencing.
Also in 5 projects that use this
Yeachan-HeoSkillultragoal
Create and execute durable repo-native multi-goal plans over Codex goal mode artifacts.
Also in 4 projects that use this
Jurel89Skillultraqa
QA cycling loop — build/lint/test, fix, retry up to 5 cycles
Also in 3 projects that use this
Other versions45 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 45
- Yeachan-Heo/oh-my-claudecode★ 41ksame name
- Yeachan-Heo/oh-my-claude-sisyphus★ 41ksame name
- Yeachan-Heo/oh-my-codex★ 34ksame name
- AgriciDaniel/claude-obsidian★ 16ksame name
- SethGammon/Citadel★ 927same name
- saltmd/salt.md★ 178same name
- ekadetov/llm-wiki★ 108same name
- saksham10arora-dotcom/everything-claude-codex★ 96same name
- skillberry-ai/cap-evolve★ 70same name
- WALKERKILLER/Look-Tongji-Notes★ 53same name
35 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository24 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- 1passwordSkill
- brainSkill
- brainstormSkill
- bridgeSkill
- claude-switcherSkill
- clear-cacheSkill
- commitSkill
- cross-artifactSkill
- ctoSkill
- diskSkill
- killSkill
- lookSkill
- parseSkill
- pbarSkill
- planSkill
- pruneSkill
- pushSkill
- reportSkill
- researchSkill
- swarmSkill
- vibe-auditSkill
- video-editorSkill
- webSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
34 rules. 0 are common, 26 are distinctive, and 8 are very specific to this project.
0
Common
Most of the 39 edited versions we read keep these.
26
Distinctive
This author's own choices. Worth a look.
8
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 56
Create the wiki root directory if it doesn't exist.
- Very specific: line 58
Create $WIKI_ROOT/README.md — the table of contents.
- Very specific: line 69
Check if a STYLE.md scaffold exists in the kit or project:
- Very specific: line 70
If $WIKI_ROOT/STYLE.md already exists, leave it alone
- Distinctive: line 71
If a scaffold source is configured (see Style Guide below), copy it
- Very specific: line 158
If neither variable is found, stop and tell the user to add CF credentials to .env.
- Very specific: line 189
If the user explicitly asks to publish uncommitted changes, build from "$WIKI_ROOT" directly.
- Distinctive: line 197
If the deploy fails with "project not found", then create it first:
- Distinctive: line 213
Do not: verify the deployment by fetching the URL.
- Very specific: line 226
Read the style guide: if $WIKI_ROOT/STYLE.md exists.
- Distinctive: line 242
If it exceeds 30 pages, read only the index and pages likely affected by recent changes.
- Distinctive: line 254
Write or update the page: Follow the style guide. Key principles:
- Distinctive: line 255
Write from the reader's perspective, not the developer's
- Distinctive: line 257
Use the project's own vocabulary — if the codebase calls it a "pipeline," don't call it a "workflow" in the wiki
- Distinctive: line 260
Start each page with a one-sentence summary of what this page covers
- Distinctive: line 270
Scan the codebase for major components: Look at:
- Distinctive: line 287
Prioritize: Don't try to fill every gap in one pass.
- Distinctive: line 297
Remove entries for deleted pages
- Distinctive: line 298
Add entries for new pages
- Distinctive: line 322
Phase 1 expands: Instead of looking at recent changes, audit the entire codebase.
- Distinctive: line 323
Read every source file's purpose, every config, every schema.
- Distinctive: line 325
Phase 3 is exhaustive: Don't prioritize — address every gap and every stale page.
- Distinctive: line 326
Create all missing pages, update all stale ones, delete all dead ones.
- Distinctive: line 328
Style consistency pass: After all content updates, re-read every page and check for consistency: terminology, voice, depth, formatting.
- Distinctive: line 331
Cross-reference check: Verify all inter-page links work.
- Distinctive: line 340
Create or update a specific page.
- Distinctive: line 343
If not, check for partial matches (e.g., "auth" matches "authentication.md").
- Distinctive: line 345
If the page exists, read it and the relevant code/config, then update it.
- Distinctive: line 347
If the page doesn't exist, assess whether it should.
- Distinctive: line 350
Update the index to reflect any changes.
- Very specific: line 376
When /wiki init runs and no STYLE.md exists, it copies from this path.
- Distinctive: line 387
Don't create placeholder pages for areas that haven't been built yet.
- Distinctive: line 389
Not every file, function, or config needs a page.
- Very specific: line 402
If .claude/kit.json has a rules."wiki" entry, read it and apply it as an additional instruction for this skill.
“Common” is measured: kept by at least 6 in 10 of the 39 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.
Files
Wiki state
!if [ -d wiki ]; then echo "wiki/ exists ($(find wiki -name '*.md' | wc -l | tr -d ' ') pages)"; [ -f wiki/README.md ] && echo "Index: wiki/README.md" || echo "Index: MISSING"; [ -f wiki/STYLE.md ] && echo "Style guide: wiki/STYLE.md" || echo "Style guide: not set"; else echo "No wiki/ directory — run /wiki init to create one"; fi
Recent changes
!git diff --stat HEAD~3..HEAD 2>/dev/null | tail -10 || echo "No recent commits"
Wiki
Maintain an organic project wiki that documents the system as it evolves. The wiki's shape emerges from the work — pages are created as topics arise, not from a fixed template. Every invocation does both targeted updates (what just changed) and a holistic scan (what's stale or missing).
Input
Optional argument: $ARGUMENTS
Dispatch
| Argument | Behavior |
| (empty) | Hybrid update — targeted updates from recent work + gap/staleness scan |
audit | Deep audit — full wiki-vs-codebase reconciliation, heavier than the default scan |
deploy | Deploy — build static HTML site and deploy to Cloudflare Pages |
init | Initialize — create wiki/ with README.md and optional STYLE.md scaffold |
<page-name> | Single page — create or update a specific page by name |
Wiki Location
The wiki lives at wiki/ in the project root by default. Override with wiki.root in
.claude/kit.json:
{ "wiki": { "root": "docs/wiki" } }Resolve the wiki root at the start of every invocation:
WIKI_ROOT=$(jq -r '.wiki.root // "wiki"' .claude/kit.json 2>/dev/null || echo "wiki")Initialize
Usage: /wiki init
- Create the wiki root directory if it doesn't exist.
- Create
$WIKI_ROOT/README.md— the table of contents. Start with:# Wiki Project documentation. Pages are created and maintained as the project evolves. ## Pages _(no pages yet)_ - Check if a
STYLE.mdscaffold exists in the kit or project:- If
$WIKI_ROOT/STYLE.mdalready exists, leave it alone - If a scaffold source is configured (see Style Guide below), copy it
- Otherwise, create a minimal default:
# Wiki Style Guide How to write and maintain this wiki. ## Audience _Define who reads this wiki and what they need from it._ ## Voice - Write in plain language - Explain what things do, not how they're implemented - Use the project's own terminology, not developer jargon ## Page conventions - One topic per page - Start each page with a one-sentence summary - Use headings to make pages scannable - Link between pages when topics connect
- If
- Report what was created.
Deploy
Usage: /wiki deploy
Build the wiki as a static HTML site and deploy it to Cloudflare Pages.
Prerequisites
wiki/must exist and contain at least one.mdfilescripts/wiki-build/must exist. The builder ships with the kit andinstall.shlinks it into the project (ininsidemode it is a copy). Changes to the builder reach every project's wiki on its next deploy.- Cloudflare credentials must be in
.env— the script checks for either prefix:CLOUDFLARE_ACCOUNT_IDorCF_ACCOUNT_IDCLOUDFLARE_API_TOKENorCF_API_TOKEN
What the builder produces
A static site: one HTML page per markdown file, a sidebar built from the index (README.md), in-browser
search, per-page heading contents and "last updated" stamps. In the sidebar, each ## Heading in the index
becomes a group, and groups start expanded. A button in the bottom-left corner collapses the sidebar on
desktop (Lucide panel-left) and expands it again (panel-right), also with Cmd+B / Ctrl+B; the choice is
remembered per browser. Tabs read Page · Wiki title.
The build runs in two steps: parse.mjs turns the pages into a bundle (wiki.json, search.json, images), then
a renderer turns the bundle into the site. The default is the kit's plain renderer (render.mjs). A project can
plug in its own with kit.json wiki.renderer — a command that is given the bundle directory and the output
directory. The bundle format is documented in the kit README ("wiki.json format").
A folder containing RESTRICTED.txt (lines of allow: person@example.com) is for those readers only. The plain
renderer leaves it out of the site altogether and says how many pages it skipped; a project's own renderer may
serve it to the people listed, using the allow lists and the Cloudflare Access settings the bundle carries
(.claude/wiki-access.json: { "team": "https://<team>.cloudflareaccess.com", "auds": [...] }, or a file passed
with --access).
Procedure
- Resolve the wiki root:
WIKI_ROOT=$(jq -r '.wiki.root // "wiki"' .claude/kit.json 2>/dev/null || echo "wiki") - Verify prerequisites:
[ -d "$WIKI_ROOT" ] && [ -d scripts/wiki-build ] || exit 1 - Load Cloudflare credentials from
.env. If the project keeps its secrets in 1Password (secrets.vaultin.claude/kit.json, values in.envwritten asop://…references), read them through the kit wrapper, which resolves references and passes plain values through:
Otherwise extract only the CF variables withCF_ACCT=$(~/.claude-kit/scripts/oprun get CLOUDFLARE_ACCOUNT_ID) CF_TOKEN=$(~/.claude-kit/scripts/oprun get CLOUDFLARE_API_TOKEN)grep(the.envmay contain non-shell content):
If neither variable is found, stop and tell the user to add CF credentials toCF_ACCT=$(grep -E '^(CLOUDFLARE_ACCOUNT_ID|CF_ACCOUNT_ID)=' .env | head -1 | cut -d= -f2-) CF_TOKEN=$(grep -E '^(CLOUDFLARE_API_TOKEN|CF_API_TOKEN)=' .env | head -1 | cut -d= -f2-).env. - Resolve the CF Pages project name. Check
kit.jsonfirst, then derive from the repo directory name:PROJECT=$(jq -r '.wiki.pagesProject // empty' .claude/kit.json 2>/dev/null) if [ -z "$PROJECT" ]; then PROJECT="$(basename "$(pwd)")-wiki" fi - Resolve the sidebar title from
kit.json(wiki.title). If it's not set, the builder uses the index's first heading:
And the renderer, if the project set one (empty means the plain renderer):TITLE=$(jq -r '.wiki.title // empty' .claude/kit.json 2>/dev/null)RENDERER=$(jq -r '.wiki.renderer // empty' .claude/kit.json 2>/dev/null) - Build from the committed pages, so uncommitted drafts are never published, running from the
project root so
.claude/wiki-access.jsonresolves:
If the user explicitly asks to publish uncommitted changes, build from(cd scripts/wiki-build && npm install --silent) SRC=$(mktemp -d) && git archive HEAD "$WIKI_ROOT" | tar -x -C "$SRC" ARGS=(); [ -n "$TITLE" ] && ARGS+=(--title "$TITLE") # an array: works in bash and zsh [ -n "$RENDERER" ] && ARGS+=(--renderer "$RENDERER") node scripts/wiki-build/build.mjs "$SRC/$WIKI_ROOT" _wiki-site "${ARGS[@]}" rm -rf "$SRC""$WIKI_ROOT"directly. - Deploy to Cloudflare Pages. Most wikis already have an existing Pages project —
skip project creation and deploy directly:
If the deploy fails with "project not found", then create it first:CLOUDFLARE_ACCOUNT_ID="$CF_ACCT" CLOUDFLARE_API_TOKEN="$CF_TOKEN" \ npx wrangler pages deploy _wiki-site --project-name="$PROJECT" --branch=main --commit-dirty=true
Then retry the deploy command above. A new project is public until it's gated: put Cloudflare Access in front of it (the production hostname andCLOUDFLARE_ACCOUNT_ID="$CF_ACCT" CLOUDFLARE_API_TOKEN="$CF_TOKEN" \ npx wrangler pages project create "$PROJECT" --production-branch=main*.<project>.pages.devpreviews) before the first deploy when the wiki shouldn't be public. - Clean up the build output:
rm -rf _wiki-site - Report the live URL:
https://<project-name>.pages.dev
Do not verify the deployment by fetching the URL. Wikis are often behind Cloudflare Access or other auth layers and cannot be reached for verification.
Hybrid Update (default)
Usage: /wiki (no arguments)
This is the core procedure — every invocation runs both halves.
Phase 1 — Context gathering
- Read the style guide if
$WIKI_ROOT/STYLE.mdexists. All writing in this invocation follows it. - Assess recent work. Gather what changed:
Also consider the current conversation — what was just built, fixed, or changed.git diff --name-only HEAD~3..HEAD 2>/dev/null git log --oneline -5 2>/dev/null - Read the wiki index (
$WIKI_ROOT/README.md) and scan existing page filenames:find "$WIKI_ROOT" -name '*.md' -not -name 'README.md' -not -name 'STYLE.md' | sort - Read all existing wiki pages. The wiki should be small enough to read in full. If it exceeds 30 pages, read only the index and pages likely affected by recent changes.
Phase 2 — Targeted updates
For each area touched by recent work:
- Does a wiki page already cover this area?
- Yes → Read it. Is it still accurate after the changes? Update if not.
- No → Should this area have a page? Not everything deserves one. A page earns its place when a reader would look for it — a major component, an integration, a workflow, a data model. Minor utilities and internal helpers usually don't.
- Write or update the page. Follow the style guide. Key principles:
- Write from the reader's perspective, not the developer's
- Explain what and why, not implementation details that change
- Use the project's own vocabulary — if the codebase calls it a "pipeline," don't call it a "workflow" in the wiki
- Link to other wiki pages when topics connect (standard markdown links)
- Start each page with a one-sentence summary of what this page covers
- Don't over-document. A wiki page that restates what the code already says is noise. Pages should explain things the code can't tell you: why something exists, how components relate, what the business context is, what decisions were made.
Phase 3 — Gap and staleness scan
Walk the project and compare against wiki coverage:
- Scan the codebase for major components. Look at:
- Top-level directories and their purpose
- Configuration files (what services, integrations, infrastructure)
- Database schemas (what data exists)
- API routes / endpoints (what the system exposes)
- Key dependencies (what external services it connects to)
- Compare against the wiki. For each major component:
- Is it documented? → Note the gap
- Is the documentation current? → Check for renamed files, removed features, changed integrations, updated schemas
- Check for dead pages. Does any wiki page describe something that no longer exists?
- Removed feature → Delete the page
- Renamed/moved feature → Update the page
- Significantly changed → Rewrite the relevant sections
- Prioritize. Don't try to fill every gap in one pass. Address:
- Stale pages first (wrong documentation is worse than missing documentation)
- Then the most important gaps (core components, integrations, data models)
- Note remaining gaps at the end for future passes
Phase 4 — Reconcile the index
Update $WIKI_ROOT/README.md:
- List every page with a one-line description
- Group pages logically (by area, not alphabetically, unless the project is small)
- Remove entries for deleted pages
- Add entries for new pages
The index is the table of contents. A reader should be able to scan it and find what they need in under 10 seconds.
Phase 5 — Report
Summarize what happened:
- Pages created (with one-line descriptions)
- Pages updated (with what changed)
- Pages deleted (with why)
- Gaps noted but not yet filled (if any)
Deep Audit
Usage: /wiki audit
A heavier version of the hybrid update. Use when you suspect the wiki has drifted significantly, or periodically as a health check.
Same as the hybrid update, but with these differences:
- Phase 1 expands. Instead of looking at recent changes, audit the entire codebase. Read every source file's purpose, every config, every schema.
- Phase 3 is exhaustive. Don't prioritize — address every gap and every stale page. Create all missing pages, update all stale ones, delete all dead ones.
- Style consistency pass. After all content updates, re-read every page and check for consistency: terminology, voice, depth, formatting. Fix inconsistencies.
- Cross-reference check. Verify all inter-page links work. Check that pages that should reference each other do.
Single Page
Usage: /wiki <page-name>
Create or update a specific page.
- Resolve the page: check if
$WIKI_ROOT/<page-name>.mdexists. If not, check for partial matches (e.g., "auth" matches "authentication.md"). - If the page exists, read it and the relevant code/config, then update it.
- If the page doesn't exist, assess whether it should. Read the relevant area of the codebase, then write the page.
- Update the index to reflect any changes.
Style Guide
The style guide at $WIKI_ROOT/STYLE.md controls how wiki pages are written. It is
optional but recommended. When present, every wiki write follows it.
The style guide itself is organic — it can be updated as the project evolves. Common things to specify:
- Audience — Who reads this wiki? Technical level? Role?
- Voice — Tone, perspective, terminology preferences
- Page conventions — Structure, heading levels, how to handle code examples
- Terminology — Project-specific terms and their definitions
- What not to document — Things that belong in code comments, not the wiki
Style Guide Scaffolding
Projects can configure a scaffold source in .claude/kit.json:
{ "wiki": { "styleScaffold": "path/to/STYLE-TEMPLATE.md" } }When /wiki init runs and no STYLE.md exists, it copies from this path. This lets
an organization maintain a standard starting point while allowing per-project evolution.
Rules
- No permission needed — Execute immediately without asking.
- Follow the style guide — If
STYLE.mdexists, every write follows it. If it doesn't, use sensible defaults (plain language, reader's perspective, one topic per page). - Organic shape — Never impose a fixed page structure. Pages emerge from the project. Don't create placeholder pages for areas that haven't been built yet.
- Don't over-document — A wiki page earns its place when a reader would look for it. Not every file, function, or config needs a page.
- Delete fearlessly — A dead page is worse than a missing page. If something was removed from the project, remove its wiki page.
- Plain markdown — No framework-specific syntax, no frontmatter, no special tooling. The wiki should be readable in any markdown viewer (GitHub, Obsidian, VS Code, raw text).
- Index is the entry point —
README.mdis always current, always scannable. A reader who only opens the index should understand what the project is and where to find what they need.
Project overrides
If .claude/kit.json has a rules."wiki" entry, read it and apply it as an additional
instruction for this skill. Absent file or key means no overrides — that is the normal case.
jq -r '.rules."wiki" // empty' .claude/kit.json 2>/dev/nullDoes 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
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.

