wix-headless-cold-start

Pass

Audited by Gen Agent Trust Hub on Nov 20, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill instructions include directives to the agent intended to override default summarization behavior, such as "Read this page in full" and "Do not use a summary." It also instructs the agent to re-fetch the raw markdown via curl if it detects that the content was initially processed by a summarizing tool.
  • [EXTERNAL_DOWNLOADS]: The skill fetches documentation, templates, and scripts from official vendor domains including www.wix.com, github.com/wix/skills, and s3.amazonaws.com.
  • [REMOTE_CODE_EXECUTION]: The skill uses npx skills@latest to download and install additional functionality and dependencies from the vendor's repository. It subsequently executes these downloaded scripts using node and npx during the bootstrap and build phases.
  • [COMMAND_EXECUTION]: The instructions utilize several shell commands for system verification and project management, including node -v, npm ping, npx, git clone, and zip.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests untrusted data from a user-provided "brief" and external site files provided by the user. These inputs are processed and used to populate site content and code.
  • Ingestion points: User prompts (the "brief"), user-provided file archives (ZIP), and external URLs referenced in the prompt.
  • Boundary markers: No specific delimiters or instructions to ignore embedded commands within user files are defined.
  • Capability inventory: The agent has access to shell execution (node, npm, npx, zip), file system modifications (wix.headless.files.patch), and network API operations via ExecuteWixAPI.
  • Sanitization: The skill does not specify security-focused sanitization or validation protocols for content extracted from user-supplied briefs or code files before processing them.
  • [DYNAMIC_EXECUTION]: The skill programmatically generates configuration files, such as js/wix/config.js, and dynamically installs module entry points at runtime based on the selected business solutions.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 20, 2026, 07:33 PM