agentleFS
EgiStrEgiStr/deckto/skills/assets-generator/SKILL.md

assets-generator

Skill0 starsChanged 50 days agoMITView on GitHub

Use when a storyline references visuals that do not exist yet — generate SVG diagrams (flow, timeline, funnel, comparison), render them to PNG, and assemble the asset manifest for pitchdeck-build. Trigger on \"generate assets\", \"create the diagram\", \"make flow.png\", when design-spec references assets/ paths, or when assets-generator is invoked by pitchdeck-build for visual_source: generate. Do NOT use for assets marked existing (collect those first) or research (search instead).

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read storyline.md. For each slide, pull visual, visual_source, and any asset path from design-spec.md.

    line 28
  2. Very specific:

    If soffice is unavailable the command reports available: false — the deck build must then swap the image slide for a native shape/text …

    line 45
  3. Distinctive:

    No text-overlap, no orphan labels.

    line 65
  4. Distinctive:

    If no licensable image exists → fall back to a Phase 2 generated visual and note the substitution.

    line 76

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Assets Generator — visuals from the storyline

Generate the missing visuals for visual_source: generate slides, render them to usable PNGs, and hand pitchdeck-build a manifest. Research-needed assets go through the research path (below); existing assets are collected, not generated.

Core principle: The storyline already decided what the visual must show — this skill only decides how. Never re-design the message; render it.

When to Use

  • storyline.md has slides with visual_source: generate and no file in assets/
  • usecase-flow.md exists but assets/flow.png doesn't
  • pitchdeck-build hits a missing asset reference mid-build

Do NOT use:

  • visual_source: existing → the asset should be in assets/ already; report if missing
  • visual_source: research → web search + download path (Phase 3)
  • Visual is a data chart → build it as a native chart in the deck (pptx skill), not as an image

Phase 1 — Scan and split

Read storyline.md. For each slide, pull visual, visual_source, and any asset path from design-spec.md. Split into three buckets:

BucketAction
existingverify file exists in deck/<slug>/assets/; missing → report, do not invent
generatePhase 2
researchPhase 3

Phase 2 — Generate

Format: author as SVG (text-based, diffable, palette-driven), render with:

npx deckto assets render deck/<slug>/assets/flow.svg   # → flow.png via soffice

If soffice is unavailable the command reports available: false — the deck build must then swap the image slide for a native shape/text layout (announce this to the user; never silently ship a missing image).

Diagram recipes (pick by what the storyline says, not by taste):

RecipeUse forRule
Node flowprocess, how-it-works≤ 6 nodes; 1–2 words per node
Timeline / gatesroadmap, GTM phasesevery gate carries its unlock metric
FunnelTAM→SAM, activationnumbers flow down; no decorative arrows
Comparison columnsus/them, before/aftersame row count both sides
Swimlanemulti-actor value exchangelane per actor, arrows labeled with the verb

Design rules (all mandatory):

  1. Palette comes from design-spec.md only — dominant field, accent on the one key element. Validate text pairs with npx deckto theme contrast <fg> <bg> — must pass.
  2. Label font size ≥ fonts.minBodyPt equivalent (14pt at export scale — SVG text is sized in px; export at ~96dpi equivalent so px ≈ pt).
  3. One motif carried in (rounded corners, the rule, the chip style) — same as the deck.
  4. No text-overlap, no orphan labels; test by opening the rendered PNG.
  5. Look at the PNG you rendered (read_image) before declaring it done. A garbled render is a failed asset, not a caveat.

Phase 3 — Research bucket

For visual_source: research (photos, reference charts):

  1. Web search for the subject; prefer CC0/public-domain/CC-BY sources; record the source URL + license in the manifest — unlicensed is unshippable.
  2. Download to deck/<slug>/assets/, name by slide (s07-market.png).
  3. If no licensable image exists → fall back to a Phase 2 generated visual and note the substitution.

Never generate fake data images (a made-up chart that looks like evidence). Data images come from real numbers in the storyline or they don't come at all.

Phase 4 — Manifest

Append to deck/<slug>/assets/manifest.json:

{ "slide": 4, "file": "assets/flow.png", "source": "generated", "recipe": "node-flow",
  "sourceUrl": null, "license": null, "alt": "Capture → Diagnose → Act, three steps" }

Research entries carry source URL + license. Then hand to pitchdeck-build.

Common Mistakes

MistakeConsequence
Rendering without reading the PNG backbroken/blank assets ship to build
Generic stock photo for a data slideviolates insight rule — the visual lies
Label font below floorQA catches it late or audience can't read it
Inventing numbers for a funnelfabrication — pull from storyline or cut the element
Ignoring the available: false pathbuild references an image that doesn't exist

Honesty note

Diagram recipes are [C] practitioner convention. The contrast and font rules are the same gates the rest of deckto enforces (computed / config-driven), so this skill adds no new thresholds of its own.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.