The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run shell commands to trigger data collection via external skills and to execute its own scripts/generate.mjs utility for theme generation.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data sourced from external URLs.
Ingestion points: The generate.mjs script ingests theme.json and meta.json from the user-specified input directory, which contains data scraped from the web.
Boundary markers: The skill lacks boundary markers or specific instructions to the agent to disregard instructions that might be embedded within the scraped design tokens.
Capability inventory: The processing script has file system capabilities including reading, writing, and copying files (fs.readFileSync, fs.writeFileSync, fs.mkdirSync, fs.copyFileSync).
Sanitization: There is no sanitization or validation of the input JSON values before they are interpolated into the DESIGN.md and globals.css templates, which could allow a malicious site to inject content into the generated design system.

generate-theme