$wiki
Project knowledge base that accumulates across sessions — stored as markdown pages under .omx/wiki/ with search, add, and lint capabilities
$wiki maintains a markdown knowledge base under .omx/wiki/. Each page is a plain markdown file with tags and categories. Discoveries made in one session become searchable context in future sessions, reducing the cost of re-exploring the same parts of a codebase.
When to use it
- When saving architecture decisions, debugging findings, or recurring patterns for future sessions
- When checking what the project wiki already knows before starting an investigation
- When ingesting external docs, specs, or research into the project knowledge base
- When checking for broken links, missing tags, or stale pages
Trigger keywords: "wiki add", "wiki query", "wiki lint", "add to wiki", "search wiki"
How to invoke
codex
> wiki add "Auth uses JWT with 15-minute access tokens and 7-day refresh tokens stored in httpOnly cookies"codex
> wiki query "authentication flow"codex
> $wiki lintHow it works
Provides 5 core operations.
wiki_add— creates a new page with title, content, tags, and a category. Categories: architecture, decision, pattern, debugging, environment, session-log, reference, convention.wiki_query— searches pages by keyword and tag matching and returns ranked results. No vector embeddings, so specific terms yield the best results.wiki_ingest— accepts external docs or spec files and splits them into categorized pages.wiki_lint— checks for broken[[page-name]]internal links, missing required tags, and stale pages.wiki_read,wiki_list,wiki_delete— direct page management.
All pages are stored as markdown files under .omx/wiki/, with a shared index maintained at .omx/wiki/index.md. Fully portable and inspectable without tooling, travels with the repository.
Outputs
.omx/wiki/{slug}.md— individual knowledge pages.omx/wiki/index.md— auto-maintained index of all pages with tags and categories.omx/wiki/log.md— append-only log of add, update, and delete operations- Lint report listing broken cross-references, missing tags, and stale pages
Related skills
$deepsearch— deep codebase search whose findings are worth saving in the wiki$trace— session execution record that can be ingested as a session-log wiki page