/wiki

Project knowledge base — living, cross-referenced, always current. Initialize, ingest new knowledge, query existing, lint for staleness.

The wiki answers: "What is the current truth about X in this project?" Unlike journal entries (which are append-only historical narrative) or settled-decision docs (which are static records), wiki pages are actively maintained — when knowledge changes, the old fact is replaced, not stacked alongside.

Ships as a global skill in core.

Where the wiki lives

.claude/wiki/
├── index.md                ← Auto-maintained table of contents
├── {topic-1}.md            ← Knowledge pages (kebab-case, one concept per page)
├── {topic-2}.md
└── ...

Pages are also indexed inside the project's root CLAUDE.md via a <!-- wiki:index --> marker block — auto-rebuilt by /save-learnings so every Claude session loads the wiki map at start.

Four modes

init — scaffold an empty wiki

/wiki init

Creates .claude/wiki/ and an index.md template. Idempotent — re-running on an already-initialized wiki is a no-op (wiki: already initialized (N pages)).

init is also fired automatically on a project's first session when atl setup-hooks is configured but the project has no .claude/wiki/ yet.

ingest — pull knowledge from project sources into wiki pages

/wiki ingest

Scans every project knowledge source and updates wiki pages:

Source	Why it matters
<!-- learning --> markers in the current session transcript	Primary source — live conversation knowledge
.claude/journal/*.md	Per-agent, per-date learning record (the post-Q4 single layer)
.claude/docs/*.md	Settled-decision documents from completed brainstorms
.claude/brain-storms/*.md (status: completed only)	Decision context
Recent conversation context	What was just discussed or built

For each piece of knowledge: determine the topic, find or create the wiki page, merge new info while resolving contradictions, update cross-references and index.md.

Wiki pages reflect current truth. If old memory says "we use pattern X" but later memory says "X caused problems, switched to Y," the wiki page says "we use Y" — not both.

query — search the wiki for a question

/wiki query how does caching work in this project?

Reads index.md to find relevant pages, reads those pages, synthesizes an answer with citations to the source pages.

Useful when:

• A new developer (or a new Claude session) needs to understand a topic without re-reading the whole repo
• You forgot how something was decided / implemented
• You want a quick refresher before making a change

lint — health check the entire wiki

/wiki lint

Checks performed:

Check	Action
Stale pages (>30 days, source files have changed since)	Flag for review
Contradictions between pages	Flag for resolution
Orphan pages (no incoming links)	Suggest connections
Missing pages (referenced but no file)	Suggest creation; create stub
Duplicate topics	Suggest merge
Index sync (index.md ↔ actual files)	Auto-fix if out of sync

Auto-fixable issues are fixed silently. Contradictions and stale content are reported for human review.

Sample output:

Wiki Health Report:
──────────────────────────
📊 Total pages: 12
✅ Healthy: 9
⚠️  Stale (>30 days): 2 (caching-patterns.md, email-setup.md)
❌ Contradiction found: auth.md says "15min token" but jwt-config.md says "30min token"
🔗 Orphan: database-indexes.md (no incoming links)
📝 Missing: "rate-limiting" referenced in api-endpoints.md but no page exists
──────────────────────────

Fixing automatically...
✅ index.md synced
✅ Created rate-limiting.md (stub)
⚠️  Review needed: auth.md vs jwt-config.md contradiction
⚠️  Review needed: 2 stale pages

Page format

Wiki pages follow a consistent structure so both humans and agents can read them quickly:

# {Topic Title}

> Last updated: {date}
> Sources: [journal](../journal/...), [brainstorm](../brain-storms/...)

## Summary
{2-3 sentence overview}

## Current State
{What is true RIGHT NOW — not history, not plans, just current reality}

## Key Decisions
{Important decisions about this topic, with brief reasoning}

## Patterns & Rules
{Established conventions for this topic}

## Known Issues
{Current problems or limitations}

## Related
- [{related-topic-1}]({related-topic-1}.md)
- [{related-topic-2}]({related-topic-2}.md)

How it interacts with other skills

The wiki is mostly auto-maintained — humans rarely edit pages by hand. The normal flow when atl setup-hooks is configured:

1. Claude drops <!-- learning topic=... --> markers inline during a conversation (per learning-capture rule)
2. SessionStart of the next session → atl session-start runs atl learning-capture --previous-transcripts, scans for markers
3. If markers found → Claude invokes /save-learnings --from-markers
4. /save-learnings updates journal/, agent children/, skill learnings/, and wiki pages (replace / update)

Example marker propagation:

<!-- learning topic: redis-cache; body: TTL should be 30 min, not 15 -->
  → journal/{date}_{agent}.md: append historical note with date
  → wiki/redis-cache.md: UPDATE "TTL is 30 minutes" (replace old "15 minutes")
  → (if domain-specific) agents/{agent}/children/redis-cache.md

/brainstorm done similarly ingests its decisions into the wiki when a brainstorm completes.

Without markers (or without hooks), /wiki ingest and /save-learnings (manual mode) still work — both can be invoked directly.

Important rules

1. Wiki = current truth. Not history, not plans. What is true RIGHT NOW.
2. Update, don't append. When a fact changes, the old version is replaced.
3. Cross-reference always. Every page links to related pages. Orphans are flagged by lint.
4. Auto-maintained. Humans rarely edit wiki directly — /save-learnings, /brainstorm done, and /wiki ingest keep it current.
5. Agent-readable. Pages are structured for both human and AI consumption — clear sections, no ambiguity.
6. Topic-based, not date-based. Unlike journal (date-based), wiki is organized by topic. One page per concept.
7. Lint regularly. Run /wiki lint periodically (monthly, or when something feels off).

Related

• /save-learnings — writes the wiki pages this skill queries / lints
• /brainstorm — done mode ingests decisions into wiki
• Concepts: Skill — wiki vs journal vs settled docs

Source

• Spec: core/skills/wiki/skill.md