The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied component descriptions, which creates a potential surface for indirect instructions to influence generated code.
Ingestion points: User descriptions enter the agent context via the $ARGUMENTS placeholder in SKILL.md.
Boundary markers: The skill does not employ specific delimiters or instruction sets to isolate user input from the core system prompt.
Capability inventory: The skill is configured to use Write, Edit, and Bash tools, enabling the creation and modification of files in the app/ and resources/ directories.
Sanitization: No validation or sanitization mechanisms are defined to check the user-provided description for malicious code or instruction overrides before file generation.

moonshine-component