The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local shell script at /Users/angel/figma/mcp-bridge/start.sh. This provides a mechanism for arbitrary command execution on the host system via an unverified script.
[COMMAND_EXECUTION]: The skill invokes an external Python script located at .claude/skills/ui-ux-pro-max/scripts/search.py. The safety and logic of this script are not contained within the skill and cannot be verified.
[EXTERNAL_DOWNLOADS]: The troubleshooting section suggests running npm install within a local directory, which results in the download and installation of external packages from the npm registry.
[PROMPT_INJECTION]: The import_html tool processes HTML content to create Figma nodes, which creates a surface for indirect prompt injection.
Ingestion points: The html parameter in the import_html tool call (SKILL.md).
Boundary markers: Absent. The skill does not define specific delimiters to separate untrusted HTML data from the agent's instructions.
Capability inventory: The skill has broad permissions to interact with Figma, including tools like import_html, create_frame, modify_node, set_variable, and delete_node (SKILL.md).
Sanitization: Absent. There is no evidence of input validation or sanitization for the HTML data processed by the bridge.

figma