| name | display |
| description | Expose the organized knowledge base through a readable frontend experience. Use when a user wants to browse the knowledge base visually as a lightweight site instead of reading raw Markdown files directly. Supports building viewers for specific data paths (e.g., individual books) or the entire knowledge base. |
| argument-hint | ["optional-data-path"] |
| allowed-tools | Bash(*), Read, Glob, Grep, Write, Edit |
Tapestry Display
Publish a readable frontend for the knowledge base: $ARGUMENTS
When to use this skill
Use this skill when:
- A user wants to browse the knowledge base visually
- You need to generate a readable frontend for the organized content
- The user asks to "view", "display", "publish", or "preview" the knowledge base or a specific book
- A lightweight site presentation is preferred over raw Markdown files
- The user wants a blog-like or research portal view of the content
Purpose
This skill acts as the presentation layer for the knowledge base.
It should:
- scan the specified data directory hierarchy (or default to
_data/books/)
- preserve the topic and chapter structure defined by
index.md
- generate a readable frontend that feels closer to a blog, notebook, or research portal than to a file browser
- support building viewers for specific books or the entire knowledge base
Workflow
-
Resolve the data path from the argument:
- If a specific path is provided (e.g., "markets-and-trading"), use
_data/books/markets-and-trading
- If no argument is provided, default to
_data/books/ (entire knowledge base)
-
Ensure the data directory exists and contains markdown files
-
Publish the frontend bundle with the appropriate data path:
python display/_scripts/publish_viewer.py --data-path _data/books/markets-and-trading --force
python display/_scripts/publish_viewer.py --force
- IMPORTANT: The viewer is created at
<data-path>/_viewer. Serve that generated directory directly:
python -m http.server 8766 --directory _data/books/_viewer
python -m http.server 8766 --directory _data/books/markets-and-trading/_viewer
Critical: When the goal is to display ALL knowledge bases, always build without --data-path
and serve _data/books/_viewer. Serving from a topic-specific _viewer/ will show only
that one topic. If you used --data-path by mistake, re-run publish_viewer.py --force
(without --data-path), kill the old server, and restart from _data/books/_viewer.
- Report back with:
- the data source path
- the viewer output directory
- the generated manifest path
- the local preview URL if served
Rules
- Treat the
index.md hierarchy as the authoritative structural map.
- Do not flatten the topic/chapter tree into a single undifferentiated list.
- Preserve topic-level separation so semantically distant materials remain clearly separated.
- Make documents readable first, but keep enough structural information visible for navigation.
- If the knowledge base is sparse or incomplete, still generate the frontend and let empty sections remain honest rather than faking content.
Common Issues
Viewer Shows Only One Topic Instead of All Knowledge Bases
Symptom: The viewer loads but only displays one topic (e.g., dingyi-dex-weekly) even though multiple KB topics exist under _data/books/.
Root cause: publish_viewer.py was called with --data-path _data/books/<topic>, which scopes the viewer to just that topic. The generated _viewer/ inside the topic directory is then served instead of the full KB viewer.
Solution:
python display/_scripts/publish_viewer.py --force
fuser -k <port>/tcp
python -m http.server <port> --directory _data/books/_viewer
Rule: To show ALL KB topics, always build and serve from _data/books/_viewer.
Only use --data-path when intentionally scoping to a single book.
Data Path Not Found
The generated viewer expects data/knowledge-base.json under <data-path>/_viewer/. If you see a JSON parsing error like "Unexpected token '<'", it usually means _ui/ was served directly instead of the published _viewer/ output.
Solution: Re-run publish_viewer.py and serve <data-path>/_viewer.
Resources
display/_scripts/publish_viewer.py: scans _data/books/, copies frontend assets, and generates the JSON manifest.
_ui/: custom static frontend assets for the viewer.