zentable（zentable 表格）

ZenTable Skill

Render structured table data into high-quality PNG images.

TL;DR

ZenTable turns messy table-like content into readable, decision-ready table outputs for both mobile (Discord-first) and desktop.

Accepted inputs

• text-based content (raw text tables, long agent responses)
• structured JSON
• screenshots / real-world photos (via OCR-assisted extraction when needed)

Core capabilities

• CSS + PIL rendering
• sorting / filtering / pagination
• threshold-based highlighting
• optional PNG + TXT dual output (--both)

Runtime & security note

This skill runs local scripts and depends on local runtime binaries (python3, google-chrome). Review code and dependencies before running in sensitive environments.

Naming convention

• Canonical code name: zentable (lowercase)
• UI / brand label: ZenTable
• zeble* / zenble* are legacy compatibility aliases
• Reference: NAMING_MIGRATION.md (repository document)

When to use

✅ Use this skill when:

• You need a visual table image instead of plain text
• You need polished output for chat/report/social sharing
• The dataset is large enough that plain text is hard to read
• You need a specific visual theme (iOS-like, dark, compact, etc.)

❌ Do not use this skill when:

• The table is tiny and plain text is enough
• The user explicitly asks for no image output
• The user needs an editable spreadsheet format (CSV/Excel)

Capability matrix (SkillHub release)

Known limitations

• ASCII output is sensitive to platform font fallback and whitespace handling.
• Create separate calibration profiles per platform; do not share blindly.
• --both already includes text-theme fallback to default when no text theme exists.
• Discord plain text collapses repeated normal spaces; Unicode spacing characters may be required for spacing preservation.
• This beta was validated primarily on Discord; other chat platforms may require agent-side output adaptation (image/message formatting differences).

Zx shorthand policy (project rule)

When user input is Zx, treat it as a strong render intent:

1. Execute rendering directly by default (no preliminary Q&A).
2. Default path: CSS + minimal_ios_mobile + width=450.
3. Ask follow-up only under high uncertainty:

- no usable source data in current/previous context, - intent does not look like table rendering, - critical fields are missing and output would likely be wrong.

1. If platform supports images, return the image directly (not link-only).

Data source priority for Zx: 1) current message image OCR 2) current message text-to-table 3) previous message image OCR 4) previous message text-to-table

Syntax sugar → canonical mapping

Pinned default baseline:

• theme=minimal_ios_mobile
• width=450
• smart_wrap=true
• per_page=15

page_spec rules

• N: page N only
• A-B: inclusive range A..B
• A-: from A to last page
• all: all pages
• if omitted: default preview pages 1-3

Canonical payload example

{
  "theme": "minimal_ios_mobile",
  "width": 900,
  "transpose": false,
  "keep_theme_alpha": false,
  "per_page": 15,
  "page_spec": "2-",
  "sort_spec": "score:desc,name:asc",
  "sort_default_dir": "asc",
  "filters": ["col:!note,attachment", "row:status!=disabled;score>=60"],
  "text_scale": "auto",
  "smart_wrap": true,
  "output_both": false
}

Command examples

# run from repository root

# basic CSS output
python3 skills/zentable/table_renderer.py - /tmp/out.png --theme minimal_ios_mobile --width 900 --text-scale large --page 1

# transpose + disable smart wrap
python3 skills/zentable/table_renderer.py - /tmp/out.png --theme compact_clean --transpose --no-smart-wrap --page 1

# page range expansion (2-4)
python3 skills/zentable/table_renderer.py - /tmp/out.p2.png --per-page 12 --page 2
python3 skills/zentable/table_renderer.py - /tmp/out.p3.png --per-page 12 --page 3
python3 skills/zentable/table_renderer.py - /tmp/out.p4.png --per-page 12 --page 4

# PNG + ASCII side output
python3 skills/zentable/table_renderer.py - /tmp/out.png --theme mobile_chat --both

Validation checklist (minimum)

• python3 -m py_compile scripts/zentable_render.py
• CSS smoke output succeeds
• PIL smoke output succeeds
• --pin, --pin-reset, --both verified
• Golden tests pass when relevant

Release positioning

Current channel: beta. ASCII remains beta/experimental for SkillHub release.

Support / Contact

• GitHub Issues: https://github.com/con2000us/zenTable/issues
• Maintainer: @con2000us (Discord)
• Bug report checklist:

- input type (text / screenshot / photo / json) - expected vs actual output - platform (Discord/mobile/desktop) - command/options used