The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the project's root directory via DESIGN-BRIEF.md to extract design specifications.
Ingestion points: DESIGN-BRIEF.md is read to determine primitive token values.
Boundary markers: The instructions do not define delimiters or warnings to ignore instructions embedded within the brief.
Capability inventory: The skill has access to shell execution (find, grep, node), file system write access, and network capabilities via curl and Playwright.
Sanitization: There is no evidence of filtering or sanitizing the content of the design brief before use.
[DYNAMIC_EXECUTION]: The skill generates and executes a local JavaScript file (_tokens_screenshot.mjs) at runtime to facilitate UI screenshots using Playwright.
[COMMAND_EXECUTION]: Extensive use of shell utilities like grep and find is employed to audit the codebase for design values, and curl is used to monitor local development server status.
[DATA_EXPOSURE]: The skill attempts to read authentication state JSON files (e.g., uiux-auth-state.json) from the /tmp directory to capture screenshots of restricted application routes.

tokens