The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze content from external websites (e.g., Dribbble, Behance) using multimodal AI. This creates a surface for indirect prompt injection where instructions embedded within images or web pages could influence the agent's behavior.
Ingestion points: Captures and analyzes screenshots from third-party inspiration websites using chrome-devtools and ai-multimodal analysis.
Boundary markers: The instructions do not define specific delimiters or warnings to ignore potential commands found within the analyzed visual assets.
Capability inventory: The agent has capabilities for file system writes (./docs/), shell execution (bash for gallery-dl), and web navigation.
Sanitization: No explicit sanitization or filtering of text extracted from images is described.
[COMMAND_EXECUTION]: The skill mentions using the shell to execute asset management utilities.
Evidence: references/design-resources.md suggests using gallery-dl via bash to download reference images.
Context: This is a routine operation for a design-focused skill to fetch assets from well-known community platforms.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading assets from established design galleries.
Evidence: Mentions using gallery-dl to fetch images from sites like Dribbble, Mobbin, and Behance.
Context: The downloads target well-known design services and are necessary for the skill's stated purpose of visual research.

aesthetic