The Agent Skills Directory

[PROMPT_INJECTION]: The section titled 'FINAL STEP' contains instructions that simulate a previous user interaction by claiming the user has already expressed dissatisfaction ('The user ALREADY said "It isn't perfect enough..."'). This technique is used to bypass normal execution flow and force the agent into a specific refinement loop by manipulating its perception of the conversation history.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to 'Download and use whatever fonts are needed to make this a reality.' This encourages the agent to fetch arbitrary external resources from the internet without specifying trusted domains or providing integrity verification for the downloaded assets.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied instructions to create a design philosophy and artwork, creating an attack surface where malicious instructions could be embedded in the input data.
Ingestion points: User-provided 'subtle input' and niche references enter the agent's context in the 'DESIGN PHILOSOPHY' and 'DEDUCING THE SUBTLE REFERENCE' sections (SKILL.md).
Boundary markers: Absent. There are no delimiters or specific instructions for the agent to treat user input as untrusted data or to ignore embedded instructions.
Capability inventory: The skill allows the agent to create and save files (.md, .pdf, .png) and potentially download external content.
Sanitization: Absent. External content is interpolated directly into the creative process and the resulting artifacts without filtering or escaping.

canvas-design