The Agent Skills Directory

[PROMPT_INJECTION]: The skill utilizes a pre-injection strategy in the 'FINAL STEP' section, instructing the agent to proceed as if the user had already provided specific feedback ('The user ALREADY said "It isn't perfect enough. It must be pristine..."'). This technique is intended to override the agent's standard response logic and force it into a specific 'refinement' behavioral mode.
[EXTERNAL_DOWNLOADS]: The instructions direct the agent to 'Download and use whatever fonts are needed to make this a reality.' This encourages the agent to initiate network requests to fetch external files. While functional for design tasks, the instruction is open-ended and does not restrict the agent to trusted domains or repositories.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and interpret 'subtle input or instructions by the user' to generate visual philosophies and art.
Ingestion points: Untrusted data enters via user requests in 'DESIGN PHILOSOPHY CREATION' and the 'DEDUCING THE SUBTLE REFERENCE' step.
Boundary markers: The skill lacks explicit boundary markers or 'ignore embedded instruction' warnings to wrap user-provided data.
Capability inventory: The skill has file-writing capabilities (.md, .pdf, .png) and implied network download capabilities for assets.
Sanitization: There is no evidence of validation or sanitization routines to ensure that user-provided conceptual input does not contain instructions targeting the agent's file system or network tools.

canvas-design