data-infographic

Turn data, comparisons, processes and timelines into explanatory visuals with Canvora — process flows, comparison charts, timelines, stat cards, checklists and long-form infographics. Use when the user has numbers, steps, a before/after, a roadmap, or asks for an infographic or diagram.

Infographics and explanatory visuals

The format is decided by the shape of the information, not by the platform. Get that mapping right and the output is good on the first run. These formats are the least discoverable in the catalog and among the most useful.

Setup and credits: see the canvora skill.

Shape → format

The information is…	Format	Size
A sequence of steps	process_flow	3:4 1200×1600
A vs B, or before/after	comparison_chart	4:3 1200×900
Events over time, a roadmap	timeline	3:4 1200×1600
A mix of stats and points	infographic	4:5 1080×1350
A long scroll of many points	infographic_long	2:3 1080×1620
A list to tick off	checklist	4:5
One number that matters	stat_card	1:1
Two options side by side, social-sized	comparison_card	4:5
Branching, non-linear relationships	mind_map	4:3 1600×1200
A single how-to tip	tip_card	1:1

Rule of thumb: linear → process_flow; two-sided → comparison_chart; dated → timeline; one number → stat_card; everything else → infographic.

Intake

1. Show me the data. Ask for the actual numbers, steps, or rows — not a description of them. This is the one skill where vagueness in equals garbage out.
2. What's the takeaway? One sentence. It becomes the headline and decides what gets emphasised.
3. Where does it go? Social → the 4:5 formats. Deck or doc → the 3:4/4:3 ones. Both → generate both, they're 10 credits each.
4. Brand kit → --brand.

The canonical runs

bash

# Steps → process flow, wording preserved
canvora generate --input "$STEPS_TEXT" --exact \
  --format process_flow --brand $BRAND --wait --json

# Us vs them
canvora generate --input "$COMPARISON_TABLE" --exact \
  --format comparison_chart,comparison_card --brand $BRAND --wait --json

# The one stat from a report
canvora generate --input "73% of returns happen in the first 48 hours" --exact \
  --format stat_card --brand $BRAND --logo --wait --json

# Full infographic from a report PDF
canvora generate --file-url $REPORT_URL --format infographic_long --brand $BRAND --wait --json

# Roadmap
canvora generate --input "$MILESTONES" --exact --format timeline --brand $BRAND --wait --json

These are all single-image formats at 10 credits each — tell the user the total in the same message as the run, announcing it rather than asking permission. --wait is fine at 1-2 outputs. If you batch 3 or more formats into one call, drop --wait, capture the id with ID=$(canvora generate ... | head -1) and poll canvora status "$ID" --json instead — most harnesses kill a shell command at 120s while the generation carries on server-side. Never re-run generate after a timeout; that starts a second, separate job. See the canvora skill's "Long runs" section.

The rule specific to data: use --exact

Numbers, labels, step names and dates must survive the pipeline unchanged. --input "..." --exact disables rewording. A rephrased statistic is a wrong statistic — this matters more here than anywhere else in Canvora.

Related: don't ask the model to derive data. If the user says "make a chart of our growth", get the actual figures first. Canvora renders what you give it; it is not a spreadsheet.

Craft notes

• Keep labels short. ≤6 words per step/row. Long labels are the #1 cause of cramped, unreadable infographics.
• One takeaway per visual. Two competing messages → two visuals.
• --style: data reads best as minimal or corporate; consumer-facing explainers as playful; reports as modern.
• --no-text is almost never right here — the text is the content.

Delivery

• Read the numbers back to the user against the source before they publish. A rendering error in a statistic is the worst failure mode in this skill.
• Wrong label or figure? canvora edit <outputId> --prompt "change 73% to 78%" --yes — 3 credits, design preserved.

Quality checklist

• Format matches the information's shape
• Real data supplied, never invented
• --exact used so figures and labels survive verbatim
• Labels ≤6 words; one takeaway per visual
• Numbers verified against the source before delivery