The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied brand descriptions or color scheme requests via the $ARGUMENTS placeholder to generate PHP code and modify configuration files. This creates a vulnerability surface where malicious instructions in the input could influence the agent's code generation or tool usage.
Ingestion points: External data enters the agent's context through the $ARGUMENTS variable at the end of SKILL.md.
Boundary markers: The skill lacks explicit delimiters or instructions to the agent to treat the user-provided content as data only and to ignore any embedded instructions.
Capability inventory: The skill utilizes Write and Edit tools to create new PHP classes in app/MoonShine/Palettes/ and modify the primary configuration file config/moonshine.php. It also has access to the Bash tool, providing a significant capability tier for executing generated content.
Sanitization: There is no evidence of input validation or sanitization mechanisms to filter out potentially malicious code or instructions within the user-provided arguments.

moonshine-palettes