What problem does it solve? Mermaid diagrams often contain syntax errors, unreadable labels, or text that vanishes on dark backgrounds, and authors cannot reliably judge their own output. This Skill renders a .mmd file to PNG on both light and dark backgrounds, inspects the images, and reports concrete problems without ever editing the source. ## Core Features & Use Cases - Syntax lint via render: Runs mmdc against the .mmd file so a parse error fails the render, and reports the verbatim error with the offending source line. - Dual-background visual check: Renders light and dark PNGs to catch text (titles, sequence labels, loop guards) that disappears on dark pages, usually caused by a pinned theme. - Clarity verdict: Checks for overlapping labels, reversed arrows, orphaned nodes, misplaced subgraph members, and whether a reader with no context could state what the diagram claims. - Use Case: A validator subagent spawned by the diagrams:mermaid workflow renders 'auth-flow.mmd', finds the title invisible on dark mode due to a pinned theme, and reports the fix to the authoring agent. ## Quick Start Ask the validator subagent to render and check the diagram in 'docs/architecture.mmd' and report whether it is correct and clear.
