The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from Figma nodes (text, layout, and names) via the Figma Bridge MCP. This data is external and untrusted, creating a surface for indirect prompt injection. However, the instructions emphasize granular data retrieval (Step 2: The Ladder) which reduces the volume of untrusted data in the context.
Ingestion points: Data entering via get_node, get_design_context, and find_nodes in SKILL.md.
Boundary markers: Absent; no specific delimiters or instruction-ignore warnings are specified for the Figma data.
Capability inventory: The skill utilizes file writing (save_children_json, get_screenshot), file reading (Read tool), and shell command execution (cp, grep, jq).
Sanitization: Not explicitly mentioned for the incoming Figma data.
[DYNAMIC_EXECUTION]: Step 2 suggests that if a server result overflows the context, the user/agent should read the dump file using a 'short Python snippet'. This involves generating and executing code to process local data.
[COMMAND_EXECUTION]: The skill encourages the use of standard CLI tools such as grep, jq, and cp to handle exported JSON files and move screenshots. It also mentions cwebp for image format conversion. These are standard developer operations within the context of an MCP bridge server.

figma-bridge