The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions specify that the agent should "Download and use whatever fonts are needed to make this a reality." This permits the agent to fetch external assets from arbitrary remote sources to fulfill artistic requests. While the included font license files reference well-known and trusted organizations such as Google, Vercel, and Red Hat, the general instruction allows for broader, unverified network requests.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and interpret user-supplied conceptual inputs to generate visual art and design manifesto documents, which creates a potential surface for indirect prompt injection.
Ingestion points: The agent processes user input to deduce "subtle, niche references" and conceptual threads for the artwork as described in SKILL.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between design instructions and potential malicious commands embedded within user-supplied input.
Capability inventory: The skill has the ability to write multiple file types (.md, .pdf, .png) to the local environment and potentially perform network operations to retrieve font files.
Sanitization: No sanitization, validation, or escaping procedures are defined for the user-provided conceptual data before it is processed by the agent's logic.

canvas-design