set_annotation / get_annotations fails on deeply nested instance nodes (3+ semicolons in node ID)

[kkjjjy1b]

Problem

set_annotation and get_annotations silently fail when targeting nodes
whose ID contains 3 or more semicolons (i.e., nested 3+ levels deep inside
component instances).

Node ID Depth Reference

Node ID format	Depth	Works?
FRAME_ID	0	✅
I{A};{B}	1	✅
I{A};{B};{C}	2	✅
I{A};{B};{C};{D}	3	❌ Returns null / no-op

Root Cause (suspected)

The Figma plugin code appears to use the synchronous figma.getNodeById(id),
which is deprecated and known to return null for nodes deep inside nested
instances.

The asynchronous figma.getNodeByIdAsync(id) correctly resolves nodes at
any depth and is the recommended replacement per Figma's plugin API docs.

Reference: https://www.figma.com/plugin-docs/api/figma/#getnodebyidasync

Impact

When the top-level frame is a component instance (not a plain frame),
almost all meaningful child nodes fall at depth 3 or deeper. This makes
set_annotation effectively unusable in component-heavy design files.

As a workaround, users have been wrapping instances in plain frames to
shallow the node depth — but this breaks Figma's master component sync
and slot override propagation, causing design system regressions.

Suggested Fix

Replace figma.getNodeById(id) with figma.getNodeByIdAsync(id) in the
plugin code where node lookup is performed before annotation operations.

- const node = figma.getNodeById(nodeId);
+ const node = await figma.getNodeByIdAsync(nodeId);

This should resolve the depth limitation without any breaking changes to
the existing API surface.

Environment

- Tool: TalkToFigma MCP (cursor-talk-to-figma-mcp)
- Affected tools: set_annotation, get_annotations
- Figma file structure: screens built entirely from component instances
(no plain frames at top level)


This is writed by claude