The Agent Skills Directory

[PROMPT_INJECTION]: The scripts/build_design_prompt.py script is vulnerable to indirect prompt injection by interpolating untrusted input into a generated prompt string.
Ingestion points: Data is read from the external file path provided to the --idea-json argument in scripts/build_design_prompt.py.
Boundary markers: There are no delimiters or specific instructions to the AI to ignore instructions embedded within the title, description, or category fields.
Capability inventory: The generated prompt is intended for use by an agent with Write and Edit file system capabilities; therefore, a successful injection could lead to unauthorized file modifications.
Sanitization: The script does not validate or sanitize the input strings before embedding them into the design prompt.

skill-designer