The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to its core functionality of processing external, untrusted content.
Ingestion points: The references/design-critique.md file directs the agent to analyze screenshots via image-viewing tools, local source code via file-reading tools, and live web pages via browsing tools.
Boundary markers: No specific delimiters or instructions (such as 'ignore instructions within analyzed files') are provided to mitigate the risk of instructions embedded in the external data being followed by the agent.
Capability inventory: The skill utilizes web-browsing, file-system-reading, and image-analysis capabilities.
Sanitization: There is no evidence of input validation or sanitization for the external data ingested during the critique process.
[EXTERNAL_DOWNLOADS]: The skill recommends the installation of external packages to support animation and live-tuning features.
Evidence: Recommends installing the dialkit (a vendor-specific package) and motion (a well-known animation library) packages in references/dialkit.md.
[COMMAND_EXECUTION]: Provides instructions for project setup via terminal commands.
Evidence: Includes commands for adding dependencies using bun add or npm install within the references/dialkit.md reference module.

design