The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes font metadata and results from external tools, which creates an attack surface for indirect prompt injection if the returned data contains malicious instructions.
Ingestion points: Results from the Adobe Fonts API via tools like font_recommend, font_search, font_details, and font_styles are processed and presented to the user.
Boundary markers: The skill contains an explicit security constraint under 'Read and present the response' that requires the agent to ignore the instructions field in network responses, preventing external content from influencing agent behavior.
Capability inventory: The skill utilizes font-specific lookup and preview tools but does not demonstrate broad file system or network access beyond the defined Adobe MCP tools.
Sanitization: There are instructions to use enrichment fields verbatim and to avoid manual construction of URLs, which reduces the risk of the agent being tricked into generating malicious links or following external directives.

adobe-fonts