wiki-manager

name	wiki-manager
description	Knowledge wiki awareness. Read articles, check stats, understand content lifecycle. Complements the research skill (which writes) with read and management awareness.

Wiki Management

You have access to knowledge wikis — structured collections of domain knowledge with a pipeline and lifecycle model.

Trigger Patterns

• Questions about what you know on a topic ("what do we know about X?", "check the wiki for Y")
• Follow-up after research ("how many articles did we add?", "what's the wiki status?")
• Domain questions that existing wiki articles might answer
• Requests to look up or read specific wiki content

Content Pipeline

Stage	Purpose	How content arrives
raw	Immutable source material with sha256 checksums (pipeline input, not an article tier)	research_fetch writes here
episodic	Per-source research findings, append-only (legacy tier — not actively fed by current pipeline)	Research pipeline writes here
articles	Polished, cross-linked knowledge (organized in categories)	Promoted by the user from episodic

Content flows: raw → articles (episodic tier is legacy). Your research pipeline handles raw. Articles are written by Claude with inline citations referencing raw sources.

Article Lifecycle

Articles progress through status stages: draft → verified. Articles are draft until human review confirms them. See knowledge-routing.md for the full tool routing table.

Tools

Tool	Use when	Input
wiki_search	You don't know what exists — discover articles by keyword	query, optional wiki_name
wiki_read	You know the slug — fetch full article content	wiki_name + slug, optional tier
wiki_stats	Check wiki health — how many articles per tier	Optional wiki_name

Decision Tree

Need domain knowledge?
  → Do you know the exact article slug?
      YES → wiki_read(wiki_name, slug)
      NO  → wiki_search(query) → pick slug → wiki_read(wiki_name, slug)

Checking research output?
  → wiki_stats() to see tier counts
  → wiki_read(wiki_name, slug, tier="episodic") to check specific summaries

Writing new content?
  → Use the research skill for structured research
  → Use wiki_write directly for one-off captures

Wiki vs MEMORY.md

These are different systems with different purposes:

	Wiki	MEMORY.md
Contains	Domain knowledge (facts, patterns, findings)	Relationship context (preferences, history, identity)
Scope	Shared across agents for a domain	Per-agent-group, personal
Writes	Research pipeline, wiki_write	Memory skill
Reads	wiki_search, wiki_read	Memory fragment at spawn

Don't write domain knowledge to MEMORY.md. Don't write personal context to wikis.