The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill manages its own runtime environment by automatically installing Node.js dependencies from the public npm registry. The script scripts/webcrawler/scripts/run-webcrawler.sh invokes npm install or npm ci if the dependency directory is not found.
[COMMAND_EXECUTION]: The skill automates project lifecycle tasks. The workflow command in scripts/webcrawler/src/commands/workflow.js executes shell commands defined in the project's package.json (such as build, typecheck, and push) using child_process.spawn.
[REMOTE_CODE_EXECUTION]: The skill performs dynamic module loading. In scripts/webcrawler/src/lib/fetch-page.js, the loadOptionalModule function uses the import() statement with computed paths to dynamically load the playwright library if available.
[PROMPT_INJECTION]: The skill's web scraping capabilities create an indirect prompt injection surface.
Ingestion points: Untrusted HTML content is fetched from external URLs in scripts/webcrawler/src/lib/fetch-page.js.
Boundary markers: The skill lacks explicit markers or instructions to isolate scraped content from the agent's core logic.
Capability inventory: The skill can modify sensitive project files (branding-presets.ts, .env) and execute arbitrary build scripts.
Sanitization: While the crawler parses HTML structure, it does not specifically sanitize the resulting text or metadata to prevent malicious instructions from influencing the branding proposal generation process.

storefront-branding