Agentprivacy Wiki Page

Fedwiki Page Skill

The fedwiki-page skill creates federated wiki page JSON files. Pages are stored as JSON files (no extension) in a wiki domain's pages/ folder. This skill covers file locations, story item types, ID generation, journal format, and the reindex step.

File Locations

Two separate wiki systems — write to the correct one:

Local/private wikis (served on port 3000):

~/Music/Guides/Private/{wiki-domain}/pages/{page-slug}

Public wikis (synced via Nextcloud to remote server):

~/Nextcloud/fedwiki/{wiki-domain}/pages/{page-slug}

The local server does NOT read from ~/Nextcloud/fedwiki/. Page slug = title lowercased with spaces as hyphens.

Story Item Types

Default to markdown for all prose, headings, and formatted text. Use code for shell commands and config snippets. Use paragraph sparingly — only for plain sentences needing no markup.

{"type": "markdown", "id": "a1b2c3d4e5f60001", "text": "**Bold**, `code`, [[Wiki Link]]"}
{"type": "code",     "id": "a1b2c3d4e5f60002", "text": "wiki-start"}
{"type": "paragraph","id": "a1b2c3d4e5f60003", "text": "Plain sentence."}
{"type": "assets",   "id": "a1b2c3d4e5f60099", "text": "page-slug"}

Story order: opening markdown → section headings → content → # Assets heading + assets item at the bottom.

IDs

Each item needs a unique 16-character hex ID. Generate with:

python3 -c "import uuid; print(uuid.uuid4().hex[:16])"

Never reuse IDs within a page or across pages in the same wiki.

Journal — Critical Rule

The journal must contain a create entry whose item.story is the complete story array — every item, identical to the top-level story, in the same order. Never use "story": [] and never include only the first item. This is what allows rewinding to the creation state.

"journal": [{
  "type": "create",
  "item": {
    "title": "Page Title",
    "story": [
      {"type": "markdown", "id": "a1b2c3d4e5f60001", "text": "Opening item."},
      {"type": "markdown", "id": "a1b2c3d4e5f60002", "text": "## Section"},
      {"type": "markdown", "id": "a1b2c3d4e5f60003", "text": "Every item — same order as top-level story."}
    ]
  },
  "date": 1781004226000,
  "certificate": "from marvin"
}]

certificate: "from marvin" identifies the creating agent. date is current Unix timestamp in milliseconds.

After Saving — Reindex

After writing any page file to disk, delete the cached sitemap so the wiki server rebuilds it:

FARM="$HOME/Music/Guides/Private"
DOMAIN="localhost"
rm -f "$FARM/$DOMAIN/status/sitemap.json"
rm -f "$FARM/$DOMAIN/status/site-index.json"
curl -s "http://$DOMAIN/system/sitemap.json" > /dev/null
echo "Reindexed $DOMAIN"

The wiki server must be running for the curl step to trigger a rebuild.

Checklist

• Slug matches title (lowercase, hyphenated)
• Correct pages/ folder (local vs public wiki)
• All IDs unique 16-char hex
• Default item type is markdown not paragraph
• Backtick markup on paths and commands in markdown items
• Journal create entry contains the full story array
• "certificate": "from marvin" on the create entry
• Reindexed after saving

Related Skills

• agentprivacy-wiki-journal — strip a page journal down to a clean create entry
• agentprivacy-wiki-ghost-pages — generate pages dynamically without storing them
• agentprivacy-wiki-claude-ghost — Claude-powered ghost pages via FastAPI
• agentprivacy-wiki-to-skill — export this page into a Claude SKILL.md

Verify: agentprivacy.ai · kept by the Librarian 🗃️ in the Tower's Wikis (the git-less onboarding layer) Upstream: skill.fedwiki.club/fedwiki-page-skill — vendored + re-framed; not original agentprivacy authorship