The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests design data from external Figma URLs using a Figma tool. Maliciously crafted Figma frames (e.g., using text designed to look like code or instructions) could potentially influence the agent's behavior or be copied into the generated source code. This is mitigated by a mandatory Phase 1 approval gate where the user must sign off on the proposed plan.
Ingestion points: Use of get_design_context in Phase 1a.
Boundary markers: Absent for the initial data ingestion, though the user approval step acts as a human-in-the-loop boundary.
Capability inventory: Subprocess execution (pnpm, gh, git), local file system writes, and network-based deployment (sfnext).
Sanitization: Not explicitly specified for the literal text strings copied from the design mock into component code.
[DYNAMIC_EXECUTION]: The agent dynamically generates React component source code and configuration files based on the structure and content identified in the Figma design at runtime.
Evidence: Phase 1d and Phase 2 instructions for generating index.tsx files with Page Designer decorators and metadata classes.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for project initialization, dependency management, and production-level deployment.
Evidence: Use of gh repo clone, pnpm install, and sfnext CLI commands for cartridge and MRT deployment.
[EXTERNAL_DOWNLOADS]: The skill clones external repositories and installs dependencies at runtime based on the target project.
Evidence: Instructions to use gh repo clone on a user-provided URL and pnpm install for project-specific dependencies.

figma-to-sfnext-pagedesigner