figma-comments

name	figma-comments
description	Read, post, reply to, and delete comments on a Figma file via the REST API — including pinning a comment to a specific node and threading replies. Use when the user wants to work with Figma comments programmatically — triggers: 'get Figma comments', 'read comments on this file', 'post a comment in Figma', 'leave a comment on this node', 'reply to a Figma comment', 'pin a comment to this element', 'delete a Figma comment', 'notify designers of drift', 'add a review note in Figma'. Uses the Figma REST API + a personal access token because comments are NOT reachable via the Plugin API / use_figma / the native Figma MCP.
disable-model-invocation	false

figma-comments — read, post, reply & delete Figma comments

Comments live only in Figma's REST API — the Plugin API (use_figma) and the native Figma MCP can't touch them. This skill reads comment threads, posts new comments (optionally pinned to a node), replies to existing threads, and deletes comments.

Setup — terminal + token required. This skill runs shell commands, so it works in Claude Code (including the "Code" tab inside Claude Desktop), Cursor, Codex, or Gemini CLI — it does not run in plain Claude Desktop or claude.ai chat (no shell). The Figma connector's OAuth login does not authorize these REST calls, so you must supply your own Figma personal access token: in Figma go to Settings → Security → Personal access tokens, generate one with scope File content: read (plus Comments: read/write), then set it in your shell: export FIGMA_TOKEN="figd_…". The script reads it from the environment at runtime — never put the token in a skill file.

Setup & skill boundaries

• All requests use X-Figma-Token: $FIGMA_TOKEN against https://api.figma.com.
• @mentions are a Figma UI-only feature. Putting @name in a message renders as plain text, not a clickable mention, and does not trigger a Figma notification.

Derive the file key

FILE_KEY=$(echo "$FILE_URL" | sed -E 's#.*/(design|file)/([A-Za-z0-9]+).*#\2#')

Read comments

scripts/get-comments.sh lists threads with author, message, timestamps, pinned node, and resolution state. Resolved threads are filtered out by default:

./scripts/get-comments.sh ABC123def456                 # active comments only
./scripts/get-comments.sh ABC123def456 --include-resolved
./scripts/get-comments.sh ABC123def456 --md             # request markdown message bodies

Raw endpoint:

curl -s -H "X-Figma-Token: $FIGMA_TOKEN" \
  "https://api.figma.com/v1/files/ABC123def456/comments?as_md=true"

Post a comment

scripts/post-comment.sh posts a top-level comment, a node-pinned comment, or a threaded reply:

# Plain file-level comment
./scripts/post-comment.sh ABC123def456 "Heads up: the button radius drifted from code."

# Pinned to a node (appears on that element). x/y default to 0,0 if omitted.
./scripts/post-comment.sh ABC123def456 "Padding here should be 16, not 12." --node 695:313 --x 40 --y 12

# Reply to an existing thread (get IDs from get-comments.sh)
./scripts/post-comment.sh ABC123def456 "Fixed in the latest sync." --reply 123456789

The raw POST body shape:

curl -s -X POST -H "X-Figma-Token: $FIGMA_TOKEN" -H "Content-Type: application/json" \
  -d '{"message":"Pinned note","client_meta":{"node_id":"695:313","node_offset":{"x":0,"y":0}}}' \
  "https://api.figma.com/v1/files/ABC123def456/comments"

• Pinning: include client_meta with node_id and a node_offset ({x,y} relative to the node). Figma requires node_offset whenever node_id is present — the script defaults it to (0,0).
• Replies: add comment_id (the id of the comment you're replying to) to the body. The script's --reply flag does this. Replies form a thread under the original comment.

Delete a comment

scripts/delete-comment.sh removes a comment by id (find ids via get-comments.sh):

./scripts/delete-comment.sh ABC123def456 123456789

Raw:

curl -s -X DELETE -H "X-Figma-Token: $FIGMA_TOKEN" \
  "https://api.figma.com/v1/files/ABC123def456/comments/123456789"

Notes

• A 403 on post/delete means your PAT lacks Comments: Read and write — reissue the token.
• Deleting a parent comment removes its whole thread. Delete replies first if you only want to prune part of a conversation.
• A common workflow: detect design/code drift, then post-comment.sh --node <id> to leave an actionable note pinned to the offending element so designers see it in context.