The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input, such as book titles, author names, and synopses, which are then interpolated into AI image generation prompts and passed as arguments to a local helper script. This creates a surface where malicious instructions embedded in the input could influence the agent's behavior or the output of the image generator.
Ingestion points: User-provided fields including Title & Subtitle, Author/Creator Names, and Back-Cover Content defined in the Agent Contract section of SKILL.md.
Boundary markers: The skill does not provide specific delimiters or instructions for the agent to treat user-supplied text as data only, lacking "ignore embedded instructions" warnings.
Capability inventory: The skill includes the capability to execute shell commands via a Python helper script (helpers/python/fal_image_generator.py).
Sanitization: There are no explicit instructions or mechanisms provided in the skill to sanitize or escape user-provided strings before they are incorporated into commands or prompts.
[COMMAND_EXECUTION]: The skill provides an example of a shell command to be executed by the agent or user. This command includes a --prompt argument that is intended to be filled with text descriptions. If the agent dynamically constructs this command using unvalidated user input, it presents a risk of command injection.
[DYNAMIC_EXECUTION]: The skill facilitates the runtime assembly and execution of shell commands. Specifically, it instructs the agent to use a local Python script (helpers/python/fal_image_generator.py) to generate images based on dynamically generated prompt strings.

fiction-cover-designer