The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts (scripts/get-comments.sh, scripts/post-comment.sh, scripts/delete-comment.sh) that use curl and jq to communicate with the Figma API.
[DATA_EXFILTRATION]: The skill transmits the FIGMA_TOKEN and comment data to api.figma.com. This is the intended purpose of the skill, and the destination is the official Figma API domain.
[PROMPT_INJECTION]: The skill is subject to Indirect Prompt Injection because it ingests untrusted data (comment text) from the Figma API and provides it to the agent's context.
Ingestion points: scripts/get-comments.sh fetches existing comments and thread data from Figma files via the REST API.
Boundary markers: Data is returned as structured JSON, but there are no specific markers to prevent the agent from following instructions potentially embedded within the comment messages.
Capability inventory: The agent can use the provided scripts to read, post, and delete comments on the user's behalf.
Sanitization: While the scripts use jq to handle JSON encoding safely, the textual content of the comments is not sanitized for adversarial prompt patterns.

figma-comments