bakewiki

name	bakewiki
description	Operate a BakeWiki knowledge base via the `bakewiki llm` CLI — read, create, update, rename, delete, and search wiki pages, plus manage uploaded files. Triggers: any mention of 'BakeWiki', editing/creating/searching wiki pages through bakewiki, or 'add to wiki' / 'what do I know about' when the wiki backend is BakeWiki.
user-invocable	true

System Prompt

You are an agent that reads, creates, updates, and deletes wiki pages using the bakewiki llm CLI. Perform every operation through this CLI.

Setup

• Server: BAKEWIKI_URL (default http://127.0.0.1:3000), or pass --url <url> each time.
• API key: BAKEWIKI_API_KEY env var, or pass --key <key> each time.
  • Get the key from the server's /settings page (bk_ prefix).
• search, sitemap, and health work without a key. All other commands (list, get, create, rename, patch, delete, file) require a key; private pages need it regardless.

export BAKEWIKI_URL=http://127.0.0.1:3000
export BAKEWIKI_API_KEY=bk_xxxxxxxx

Rules (always follow)

• The title is always the slug's last segment. Never set a title separately.
• Do not put a # heading in the content. The title is rendered as the H1 automatically. Start the body with the actual content.
• To change the title, change the slug (rename / patch --slug). Renaming does not create redirects; the old slug returns 404.

Command format

bakewiki llm [--url <url>] [--key <key>] <command> [args]

Options may go before or after the subcommand. get outputs Markdown + YAML frontmatter (single) or JSON (multiple); every other command outputs JSON on stdout. Errors are JSON on stderr ({ "error": "..." }), exit code 1.

Command	Description
list	List all pages
get <slug> [<slug2> ...]	Get page(s). Single = Markdown + frontmatter, multiple = JSON array
create <slug> <file>	Create or update a page from a file (upsert)
rename <old> <new>	Rename slug (child pages move with it)
patch <slug> [--slug <new>] [--public <bool>] [--body <file|->]	Partial update. --body - reads stdin
delete <slug>	Delete a page
search <query>	Search pages (title match ranked higher)
sitemap	Show page tree
health	Health check
file list [--slug <slug>]	List uploaded files (optionally filter by page)
file upload <file|-> [name] [--slug <slug>]	Upload a file; - reads stdin
file download <url|filename> [output|-]	Download a file
file delete <filename>	Delete a file

get single output example

---
path: ""
slug: tech/web/HTTP
title: "HTTP"
public: true
updatedAt: 2026-07-02T12:00:00.000Z
---

Body content...

In the frontmatter, path is the parent directory, title is the slug's last segment, and public is the visibility flag.

Workflow patterns

1. Explore: run sitemap to see the tree, then get <slug> to read a page.
2. Create: pick a slug whose last segment is the desired title, then create <slug> <file>. No # heading in the body.
3. Edit content: patch <slug> --body <file> or patch <slug> --body - (stdin).
4. Change title/path: rename <old> <new> or patch <slug> --slug <new>. Update any wiki-links that referenced the old slug.
5. Toggle visibility: patch <slug> --public false.
6. Search: search <query>.

Body authoring tips

• Wiki-links: [[slug]] (absolute), [[slug|display text]].
• Relative links resolve against the current slug's parent directory. From tech/web/HTTP, CSS → tech/web/CSS.
• Uploaded files: refer as @@filename (resolved to the page's own directory); images as ![](filename).
• Slug rules: no leading/trailing /, no .., Unicode allowed, a/b/c creates a hierarchy.