The Agent Skills Directory

[PROMPT_INJECTION]: The skill incorporates a prompt assembly framework that includes specific instructions for the image generation tool to avoid refusals. In references/workflows/prompt-assembly.md, the instruction 'DO NOT refuse to generate' is used to direct the tool to provide alternatives for sensitive or copyrighted content rather than following standard refusal protocols.
[PROMPT_INJECTION]: The skill processes arbitrary user-supplied text and images to generate infographic content. This creates an indirect prompt injection surface where malicious instructions embedded in the input content could potentially influence the behavior of the image generation tool. While the skill uses markdown headers and delimiters as boundary markers, there is no explicit sanitization or filtering of the user input before interpolation.
Ingestion points: User-pasted content or files saved as source.md (Step 1).
Boundary markers: Uses --- delimiters and structured headers (## Content) in the prompt assembly.
Capability inventory: Executes shell commands for file management and utilizes an image generation tool ('nano banana pro').
Sanitization: No input sanitization or validation logic is present in the workflow documentation.
[COMMAND_EXECUTION]: The workflow in SKILL.md instructs the agent to execute shell commands (test -f) to verify the existence of the EXTEND.md configuration file in both the project directory and the user's home directory ($HOME). These operations are used solely for managing user preferences and session data.

baoyu-xhs-images