The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted data from external websites and local repository files to synthesize design guidelines.
Ingestion points: External URLs provided by users (processed via browser automation) and local codebase files including CSS, JavaScript theme objects, and Tailwind configurations.
Boundary markers: The instructions do not define specific delimiters or warnings for the agent to ignore instructions that might be embedded in the source material.
Capability inventory: The skill has the ability to write to the file system (creating DESIGN.md), execute shell commands via npx, and perform network/browser operations.
Sanitization: While the skill uses a dedicated JavaScript script to extract structured CSS properties, prose sections such as the Overview and Do's and Don'ts are synthesized from general visual and textual signals, which are susceptible to influence from external data.
[COMMAND_EXECUTION]: The skill executes a linting utility via npx from a well-known and trusted technology organization.
[EXTERNAL_DOWNLOADS]: The skill downloads and executes a validation package from a public registry during its execution flow to ensure the generated file meets specification requirements.
[DYNAMIC_EXECUTION]: The skill uses browser automation to execute a local JavaScript utility that extracts computed styles from a webpage context.

design-md