The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided layer-tree.json files which creates a surface for indirect prompt injection if design metadata contains malicious instructions.
Ingestion points: The load_json function in scripts/split_components.py ingests external design structures.
Boundary markers: The skill lacks explicit instructions for the AI to ignore instructions embedded within the design data.
Capability inventory: The script possesses file-writing capabilities through Path.write_text and file-copying through shutil.copy2.
Sanitization: The generators in scripts/react_generator.py and scripts/vue_generator.py correctly utilize html.escape for text layers and sanitize_component_name for class names to prevent structural injection during code generation.
[DYNAMIC_EXECUTION]: The skill performs dynamic code generation, creating JSX and Vue SFC files based on design metadata.
The generation logic in scripts/react_generator.py and scripts/vue_generator.py assembles strings into executable source files.
This behavior is the primary intended function of the skill and is handled via templates and sanitization.

psd-component-splitter