The Agent Skills Directory

[SAFE]: The skill consists of instructional Markdown files that define routing logic, design parameters, and implementation constraints. Analysis of these instructions shows no attempts at prompt injection, data exfiltration, or bypassing safety filters. The logic is centered on providing high-quality UI design guidance.
[COMMAND_EXECUTION]: The evaluation harness located in experiments/eval/runner.ts utilizes child_process.execSync to execute git show commands. This functionality is used to retrieve skill file variants from different git branches for benchmarking purposes. The command arguments are derived from hardcoded configuration within the script and do not process unsanitized user input.
[EXTERNAL_DOWNLOADS]: The skill mentions external developer tools such as better-icons and utilizes standard NPM packages within its experiments directory. These dependencies are well-known libraries (OpenAI, Anthropic SDK) used for testing and asset discovery, which align with the stated purpose of the repository.

presto-design