The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a bash command (ls /mnt/skills/user/imagine/references/artstyle-*.md) to dynamically identify available art style reference files within its own directory. This allows for the extensibility of the style library and is a legitimate functional requirement for the skill's primary purpose.\n- [PROMPT_INJECTION]: The skill instructions provide guidance on configuring downstream tool parameters for the Google Imagen API, including safety settings such as allow_adult and block_none. While these instructions involve reducing safety constraints for the target image generation tool, they are documented as valid technical parameters for the intended creative workflow.\n- [SAFE]: A surface for indirect prompt injection exists where user natural language input is processed to build prompts. (1) Ingestion point: User-provided creative requests. (2) Boundary markers: Not explicitly defined in instructions. (3) Capability inventory: Listing and reading local reference files, text output; no network or file-write operations. (4) Sanitization: None documented. The risk is considered safe due to the limited capabilities and absence of high-risk operations.

imagine