Repository navigation
Agent Skill
The repository ships a Claude Code skill at .claude/skills/tayda-uv/ that teaches an agent to drive the tool. With it installed you can say "make me a Tayda PDF for the face of a 1590B from face.png, it's a black enclosure" and the agent will run sides, validate, convert and inspect in the right order, choose -white from the enclosure colour, keep gloss off unless you asked for it, and tell you to check the result with the Analyzer.
The skill encodes the rules this wiki spells out for people:
- validate before converting, and report problems rather than fixing them silently
- never pass
-forceon its own initiative - one command per side; no combining sides into one PDF
- artboard sizes come from
tayda-uv sides, never from memory -
inspectafter everyconvert, always with-eand-s - do not try to script Tayda's PDF Analyzer; that step is yours
- recommend ordering one enclosure before a batch
The skill is active automatically for any agent working inside the repository. Your artwork probably lives elsewhere, so link it into your global skills directory and make sure the binary is on PATH:
cd tayda-uv-artwork-processor
go install ./cmd/tayda-uv
ln -s "$PWD/.claude/skills/tayda-uv" ~/.claude/skills/tayda-uvAfter that any Claude Code session picks it up when the conversation mentions Tayda UV printing, pedal enclosure artwork, or artboard sizes.
It will not drive the interactive interface — that needs a real terminal, and everything it does is a flag anyway. It will not invent an artboard size, and it will not upload to Tayda for you.
The skill is a Markdown file with no Claude-specific syntax. Point any other tool-using agent at .claude/skills/tayda-uv/SKILL.md and the same instructions apply.