The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The recon-handshake command extracts sensitive browser cookies and session data from a local Chrome instance to authenticate WebSocket connections.\n- [CREDENTIALS_UNSAFE]: Extracted session data is stored by default in /tmp/figma_handshake.json, a predictable and often globally readable location, increasing the risk of credential theft by other local processes.\n- [DATA_EXFILTRATION]: The skill requires a FIGMA_TOKEN environment variable, which handles authentication for Figma's REST API and WebSocket protocols.\n- [PROMPT_INJECTION]: The skill represents an indirect injection surface by ingesting untrusted data (Figma scenegraph frames) and providing capabilities to write mutations back to the live Figma document.\n
Ingestion points: capture-all-pages command and WebSocket frame interception via Chrome DevTools Protocol.\n
Boundary markers: No explicit sanitization or instructions to ignore embedded content within captured frames are documented.\n
Capability inventory: Includes full write access via write, mutate, rename, and clone operations.\n
Sanitization: Not documented; the skill decodes binary data into JSON structures which are then processed by the agent.\n- [COMMAND_EXECUTION]: The skill relies on execution of the figma-kiwi-protocol CLI tool via npx for capture, decoding, and writing operations.

figma-kiwi-protocol-decode