The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill interpolates user-provided content from the $ARGUMENTS variable directly into its operational prompt without boundary markers or sanitization. This creates a surface where a user could provide instructions that override the skill's intended behavior, potentially abusing the agent's available tools.
Ingestion points: The $ARGUMENTS placeholder at the end of SKILL.md accepts arbitrary user input.
Boundary markers: There are no delimiters or specific instructions to the agent to treat the input as untrusted data.
Capability inventory: The skill is configured with Edit, Write, and Bash tools, which could be misused if the agent obeys instructions injected via the arguments.
Sanitization: No validation or filtering is performed on the input before processing.

moonshine-layout