The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The entry/skill.md file instructs the agent to download a bootstrap configuration script from the official wix.com domain to initialize the build environment. This resource originates from the same organization as the skill author.
[REMOTE_CODE_EXECUTION]: Instructions in entry/skill.md involve downloading and executing a JavaScript bootstrap module (bootstrap.mjs) from the vendor's domain (wix.com) to manage authentication and CLI verification. This is a part of the initial cold-start setup procedure.
[COMMAND_EXECUTION]: The installation scripts (install/fast-path.mjs, install/deploy.mjs) and seeding modules (references/*/seed/*.mjs) use Node.js child_process methods to execute standard npm commands, npx @wix/cli for token generation, and other internal deployment scripts.
[PRIVILEGE_ESCALATION]: The skill implements a server-side API endpoint for media uploads (references/shared/capabilities/media-upload/app-astro/pages/api/wix/media-upload.ts) that utilizes the platform-native auth.elevate() mechanism to generate upload URLs. This is an intended feature for authorizing specific visitor-initiated actions on the server.
[INDIRECT_PROMPT_INJECTION]: The skillRetrieved and renders content from external Wix APIs (CMS, Blog, Storefront), representing an ingestion surface for untrusted data. 1. Ingestion points: Data enters the context via API client modules like wix/cms/items.ts, wix/blog/posts.ts, and wix/storefront/catalog.ts. 2. Boundary markers: The skill uses structured DTO (Data Transfer Object) mapping to isolate raw API responses from the presentation layer. 3. Capability inventory: The skill utilizes shell command execution for installation and network requests for seeding. 4. Sanitization: The skill provides specific components like RichContent (using the Ricos viewer) to securely render complex content and encourages the use of set:html on controlled wrappers for other content types.

wix-headless-fast