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).
In this repository9 other files it gives its agents.
- llms.txtllms.txt
- audience-fitSkill
- brand-designSkill
- humanizerSkill
- pitchdeck-buildSkill
- pitchdeck-grindingSkill
- pitchdeck-pitch-meSkill
- pitchdeck-reviewSkill
- pptxSkill
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
- Very specific: line 28
Read storyline.md. For each slide, pull visual, visual_source, and any asset path from design-spec.md.
- Very specific: line 45
If soffice is unavailable the command reports available: false — the deck build must then swap the image slide for a native shape/text …
- Distinctive: line 65
No text-overlap, no orphan labels.
- Distinctive: line 76
If no licensable image exists → fall back to a Phase 2 generated visual and note the substitution.
≈ “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.
Files
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.mdhas slides withvisual_source: generateand no file inassets/usecase-flow.mdexists butassets/flow.pngdoesn'tpitchdeck-buildhits a missing asset reference mid-build
Do NOT use:
visual_source: existing→ the asset should be inassets/already; report if missingvisual_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:
| Bucket | Action |
existing | verify file exists in deck/<slug>/assets/; missing → report, do not invent |
generate | Phase 2 |
research | Phase 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 sofficeIf 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):
| Recipe | Use for | Rule |
| Node flow | process, how-it-works | ≤ 6 nodes; 1–2 words per node |
| Timeline / gates | roadmap, GTM phases | every gate carries its unlock metric |
| Funnel | TAM→SAM, activation | numbers flow down; no decorative arrows |
| Comparison columns | us/them, before/after | same row count both sides |
| Swimlane | multi-actor value exchange | lane per actor, arrows labeled with the verb |
Design rules (all mandatory):
- Palette comes from
design-spec.mdonly — dominant field, accent on the one key element. Validate text pairs withnpx deckto theme contrast <fg> <bg>— must pass. - Label font size ≥
fonts.minBodyPtequivalent (14pt at export scale — SVG text is sized in px; export at ~96dpi equivalent so px ≈ pt). - One motif carried in (rounded corners, the rule, the chip style) — same as the deck.
- No text-overlap, no orphan labels; test by opening the rendered PNG.
- 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):
- Web search for the subject; prefer CC0/public-domain/CC-BY sources; record the source URL + license in the manifest — unlicensed is unshippable.
- Download to
deck/<slug>/assets/, name by slide (s07-market.png). - 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
| Mistake | Consequence |
| Rendering without reading the PNG back | broken/blank assets ship to build |
| Generic stock photo for a data slide | violates insight rule — the visual lies |
| Label font below floor | QA catches it late or audience can't read it |
| Inventing numbers for a funnel | fabrication — pull from storyline or cut the element |
Ignoring the available: false path | build 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.
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.

