name	annotations-figma
description	Read and write designer annotations on Figma nodes — node-level design specs that pin properties (fills, width, fontSize, etc.) and carry plain or markdown notes. Use when the user wants to document or inspect interaction specs, animation timings, easing, accessibility requirements, or implementation notes attached directly to a node. Triggers: 'add an annotation', 'annotate this node/component', 'read the annotations on…', 'what specs are pinned to this element', 'list annotation categories', 'document the focus behavior on this button in Figma'. Annotations are distinct from comments (they pin to design properties) and from the description field. Requires the Figma Desktop app (Plugin API).
disable-model-invocation	false

annotations-figma — read & write node annotations

Annotations are designer-authored specs attached to a node. Each can carry a note (plain label or rich labelMarkdown), a set of pinned properties (fills, width, fontSize, …) that link the note to specific design attributes, and an optional category (interactions, accessibility, dev notes, …). They live on the node, survive edits, and are undoable (Cmd+Z). They are the right place for animation timings, easing curves, interaction behavior, and a11y requirements that don't belong in the component description.

Skill boundaries

use_figma rules — load the official figma-use skill first; it is the full Figma Plugin API reference. Essentials these scripts rely on: plain JS with top-level await + return (no IIFE, no figma.closePlugin(); console.log is not returned), inputs inlined as const at the top of each script, colors in 0–1 range, load fonts before any text op, await figma.getNodeByIdAsync(...), and atomic errors (a failed script applies nothing — read the error, fix, retry).
Valid property types + shapes → references/annotation-properties.md.
Comments (file-level discussion threads, not node specs) are out of scope — use Figma's comments REST API instead.

Workflow

Get the node ID. Use the current selection or a node ID the user gives you. Annotations only apply to nodes whose type supports them (frames, components, instances, shapes, text — not pages).
List categories first when writing categorized annotations. Run the categories snippet so you can pass a real categoryId (category names are per-file and not guessable). See the script.
Read with scripts/get-annotations.js — set INCLUDE_CHILDREN to walk a component tree for full-component documentation.
Write with scripts/set-annotations.js. Choose MODE: 'replace' (default — overwrites all annotations on the node) or 'append' (keeps existing). Pass ANNOTATIONS = [] with 'replace' to clear all annotations.
Validate. Re-run the read script and confirm the annotation count and labels match what you set.

Notes

Note any text op needs a font load. Annotations themselves don't render text you author, but if a workflow also writes text nodes, await figma.loadFontAsync(...) first (see conventions).
Figma auto-populates BOTH label and labelMarkdown on read, but rejects writing both — when appending, prefer labelMarkdown if present, else label. The append script already does this.
properties only pins which attributes the note is about; it does not set their values. Use the exact strings from references/annotation-properties.md.
These tools require the Figma Desktop app (the Plugin API isn't available in the browser).

annotations-figma — read & write node annotations

Skill boundaries

use_figma rules — load the official figma-use skill first; it is the full Figma Plugin API reference. Essentials these scripts rely on: plain JS with top-level await + return (no IIFE, no figma.closePlugin(); console.log is not returned), inputs inlined as const at the top of each script, colors in 0–1 range, load fonts before any text op, await figma.getNodeByIdAsync(...), and atomic errors (a failed script applies nothing — read the error, fix, retry).

Valid property types + shapes → references/annotation-properties.md.

Comments (file-level discussion threads, not node specs) are out of scope — use Figma's comments REST API instead.

Workflow

Get the node ID. Use the current selection or a node ID the user gives you. Annotations only apply to nodes whose type supports them (frames, components, instances, shapes, text — not pages).

List categories first when writing categorized annotations. Run the categories snippet so you can pass a real categoryId (category names are per-file and not guessable). See the script.

Read with scripts/get-annotations.js — set INCLUDE_CHILDREN to walk a component tree for full-component documentation.

Write with scripts/set-annotations.js. Choose MODE: 'replace' (default — overwrites all annotations on the node) or 'append' (keeps existing). Pass ANNOTATIONS = [] with 'replace' to clear all annotations.

Validate. Re-run the read script and confirm the annotation count and labels match what you set.

Notes

Note any text op needs a font load. Annotations themselves don't render text you author, but if a workflow also writes text nodes, await figma.loadFontAsync(...) first (see conventions).

Figma auto-populates BOTH label and labelMarkdown on read, but rejects writing both — when appending, prefer labelMarkdown if present, else label. The append script already does this.

properties only pins which attributes the note is about; it does not set their values. Use the exact strings from references/annotation-properties.md.

These tools require the Figma Desktop app (the Plugin API isn't available in the browser).

annotations-figma