visual-director
Skill1 starsChanged 3 months agoMITView on GitHub
Create a coherent, brand-aware visual direction and page-by-page layout blueprint for presentation decks. Use when a deck needs visual hierarchy, layout diversity, diagrams, data storytelling, technical architecture visuals, accessibility planning, or a polished aesthetic.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- Cheukcoding/claude-skillstheir own version, first seen Nov 1, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- Cheukcoding/claude-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository17 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- claude-code-styleSkill
- create-deckSkill
- deck-architectSkill
- deck-reviewerSkill
- diagram-designSkill
- review-deckSkill
- speaker-notesSkill
- visual-directorSkill
- claude-code-styleSkill
- create-deckSkill
- deck-architectSkill
- deck-reviewerSkill
- diagram-designSkill
- review-deckSkill
- speaker-notesSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
33 rules. About 0 are common, 28 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈28
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 10
When available, use the diagram-design skill for standalone diagram types and Mermaid/draw.io redraws.
- Very specific: line 16
Read the communication contract, Direction Lock, evidence ledger, section map, Narrative Review, and slide blueprints from deck-plan.md.
- Distinctive: line 20
Check the page against the previous two and next page for semantic fit and visual rhythm.
- Very specific: line 21
Update the layout fields in deck-plan.md when a workspace exists.
- Distinctive: line 57
Every arrow names data, identity, protocol, decision, or ownership.
- Distinctive: line 59
Every page has an internal audience takeaway, but a visible takeaway strip is default-off when the title and working area already carry the …
- Distinctive: line 66
never repeat the exact archetype on consecutive slides
- Distinctive: line 67
do not let the same visual family dominate more than two of any three consecutive slides
- Distinctive: line 68
keep card-grid and node-card pages at or below roughly 20%
- Distinctive: line 69
keep generic split-screen pages at or below roughly 35%
- Distinctive: line 70
avoid more than two consecutive slides with the same underlying geometry
- Distinctive: line 78
Default to a title-led composition and whitespace.
- Distinctive: line 79
Use the right half only for exactly one sourced artifact, compact proof rail, operational signal, or direct system cue that can be …
- Distinctive: line 80
Keep any right-side module secondary to the title and normally no wider than one-third of the canvas.
- Distinctive: line 81
Never invent generic orbit, hub-and-spoke, concentric-circle, logo-cloud, or micro-label diagrams as cover filler.
- Distinctive: line 82
If no meaningful module exists, leave the space empty.
- Distinctive: line 84
When the opening is causal or mechanism-led, keep the proof rail concise and use the next page for the mechanism.
- Distinctive: line 88
Prefer a native chart for quantitative relationships and preserve definitions, axes, units, dates, sources, and decision implications.
- Distinctive: line 89
Use a native table for repeated records with aligned columns.
- Distinctive: line 90
Use protocol lifelines for ordered interactions across actors.
- Distinctive: line 91
Use lanes for ownership and handoff.
- Distinctive: line 92
Use boundaries for trust, tenancy, network, or control-plane crossings.
- Distinctive: line 94
Use a screenshot for a real interface or artifact and a diagram for a conceptual model.
- Very specific: line 95
When a standalone architecture, flowchart, sequence, state, ER, timeline, swimlane, quadrant, loop, chart, data-flow, or security visual is …
- Very specific: line 100
Write <visual>.visual.md before <visual>.py, then generate <visual>.svg.
- Distinctive: line 101
Default to a deck-local generator for a source-specific mechanism, architecture, journey, boundary, or scene.
- Distinctive: line 103
Keep title, explanation, source note, footer, and speaker notes editable outside the SVG whenever possible.
- Distinctive: line 104
Treat built-in kinds as examples and fast fallbacks, not a closed catalog.
- Distinctive: line 105
Verify accessible metadata, deterministic regeneration, source fidelity, and readability at presentation and thumbnail scale.
- Distinctive: line 109
When using the Claude Code-inspired system, use terracotta sparingly.
- Distinctive: line 110
Prefer whitespace, hairlines, direct labels, and meaningful geometry over card grids.
- Distinctive: line 112
Never request or fabricate an official Anthropic or Claude logo.
- Distinctive: line 113
Avoid gratuitous gradients, glassmorphism, emoji, stock illustrations, and excessive rounded rectangles.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Direct the presentation visually
Read brand-neutral visual quality, the layout system, and the Python SVG authoring protocol, resolved relative to this skill directory. Use the bundled Python SVG plan template and scripts/generate-slide-art.py only when the visual workflow needs them. Read the Claude Code-inspired visual system only when the selected template is claude-editorial / terminal-editorial, or the user explicitly requests Claude styling.
When available, use the diagram-design skill for standalone diagram types and Mermaid/draw.io redraws; its selected type reference and self-check govern that visual asset. Inspect existing brand assets, screenshots, charts, deck files, and deck-plan.md. Do not rewrite the deck's narrative or silently change claims.
Resolve bundled scripts to absolute paths from the directory containing this SKILL.md, and execute them while keeping the user's project as the shell working directory. Do not change into the installed skill directory.
Workflow
- Read the communication contract, Direction Lock, evidence ledger, section map, Narrative Review, and slide blueprints from
deck-plan.md. - For each slide, identify the information shape before choosing a visual form or layout archetype.
- Select the visual form that explains the evidence-to-claim relationship fastest.
- Resolve the archetype, geometry, dominant element, density, and eye path.
- Check the page against the previous two and next page for semantic fit and visual rhythm.
- Update the layout fields in
deck-plan.mdwhen a workspace exists; otherwise return a complete layout blueprint to the orchestrator. - Audit the sequence at thumbnail scale before implementation.
Return
- A one-sentence visual concept.
- Palette, typography, grid, spacing rhythm, and motion policy.
- A page-by-page
layoutBlueprintcontaining:- slide role and claim title
- evidence status
- information shape
- selected visual form
- layout archetype
- geometry
- dominant element
- density
- intended eye path
- composition and supporting elements
- source assets and editable outside content
- transition handoff
- For each cover or section opener, a
coverRightdecision:none,artifact-right,proof-rail,signal-stack, ordirect-system-cue, with the source and why it earns its place. - For each proposed custom Python SVG, a
pythonSvgPlan: plan path, script path, output path, communication job, source of truth, semantic model, geometry, eye path, editable outside content, and validation requirements. - A layout-diversity audit: unique archetype count, consecutive repeats, visual-family rhythm, card-page share, split-page share, same-geometry runs, and planned rhythm changes.
- Diagram, native-table, chart, screenshot, code, or protocol-sequence specifications where relevant.
- Required assets and safe fallback treatments.
- Accessibility and export risks.
Semantic layout rules
- One dominant visual idea per page.
- Choose layouts because they match the page's information shape and semantic job, not to create random variety.
- A protocol interaction uses actors, message order, and labeled request/response arrows; it is not a generic architecture diagram.
- A causal chain is not automatically a timeline.
- Column-aligned records are a native table, not a matrix made from text boxes.
- Screenshots state what they prove and use no more than four callouts.
- Architecture shows real services, stores, flows, ownership, and boundaries; never add components for visual complexity.
- Every arrow names data, identity, protocol, decision, or ownership.
- Cards are an implementation tool, never the default answer to a list.
- Every page has an internal audience takeaway, but a visible takeaway strip is default-off when the title and working area already carry the implication.
Deck-level rhythm
For a deck of 10 or more slides:
- target at least eight distinct archetypes when the content supports them
- never repeat the exact archetype on consecutive slides
- do not let the same visual family dominate more than two of any three consecutive slides
- keep card-grid and node-card pages at or below roughly 20%
- keep generic split-screen pages at or below roughly 35%
- avoid more than two consecutive slides with the same underlying geometry
- introduce a visible rhythm change every three to four slides through scale, density, background, or dominant visual
- interleave claims with evidence, systems, sequences, risks, and decisions
Rhythm must remain causal. Do not place a layout merely because the deck has not used it yet.
Cover rules
- Default to a title-led composition and whitespace; the right half is optional.
- Use the right half only for exactly one sourced artifact, compact proof rail, operational signal, or direct system cue that can be understood within two seconds.
- Keep any right-side module secondary to the title and normally no wider than one-third of the canvas.
- Never invent generic orbit, hub-and-spoke, concentric-circle, logo-cloud, or micro-label diagrams as cover filler.
- If no meaningful module exists, leave the space empty.
- Give every proof-rail number a semantic role, source, date, and explicit relationship to the other signals.
- When the opening is causal or mechanism-led, keep the proof rail concise and use the next page for the mechanism. The handoff should move from evidence to causal model to implication instead of expanding the cover into a dashboard.
Diagram and data rules
- Prefer a native chart for quantitative relationships and preserve definitions, axes, units, dates, sources, and decision implications.
- Use a native table for repeated records with aligned columns.
- Use protocol lifelines for ordered interactions across actors.
- Use lanes for ownership and handoff.
- Use boundaries for trust, tenancy, network, or control-plane crossings.
- Draw a loop only when feedback changes the next cycle.
- Use a screenshot for a real interface or artifact and a diagram for a conceptual model; do not redraw evidence into a cleaner but less truthful abstraction.
- When a standalone architecture, flowchart, sequence, state, ER, timeline, swimlane, quadrant, loop, chart, data-flow, or security visual is needed, prefer
diagram-designor follow its type-specific grammar and self-check.
Custom Python SVG
- Plan from the semantic model, not from a fixed drawing kind.
- Write
<visual>.visual.mdbefore<visual>.py, then generate<visual>.svg. - Default to a deck-local generator for a source-specific mechanism, architecture, journey, boundary, or scene.
- State why Python SVG communicates better than a screenshot, verified chart, native editable shapes, or whitespace.
- Keep title, explanation, source note, footer, and speaker notes editable outside the SVG whenever possible.
- Treat built-in kinds as examples and fast fallbacks, not a closed catalog.
- Verify accessible metadata, deterministic regeneration, source fidelity, and readability at presentation and thumbnail scale.
Brand and craft
- When using the Claude Code-inspired system, use terracotta sparingly; it is emphasis, not decoration.
- Prefer whitespace, hairlines, direct labels, and meaningful geometry over card grids.
- Terminal chrome is a supporting motif, not the entire layout.
- Never request or fabricate an official Anthropic or Claude logo.
- Avoid gratuitous gradients, glassmorphism, emoji, stock illustrations, and excessive rounded rectangles.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

