The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions direct the agent to download a JavaScript utility file (utils.cjs) from an official Wix domain (https://www.wix.com/skills/wix-base44-connector/scripts/utils.cjs) and execute it using require. This is a vendor-authorized operation used to load necessary helper functions into the execution environment.
[COMMAND_EXECUTION]: The wx.bash helper function in the scripts/utils.cjs file provides the agent with the ability to execute shell commands using child_process.execSync. This is used to perform text processing (e.g., grep, sed, mawk) on documentation files stored in the local temporary directory.
[DYNAMIC_EXECUTION]: The wx.spec function allows the agent to send JavaScript code snippets to a remote Wix service (https://mcp.wix.com/api/code-mode/search) for evaluation. This mechanism is used to dynamically query and parse API schemas during the discovery phase.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface because it ingests and processes external data (API documentation and site context) from dev.wix.com and Wix APIs.
Ingestion points: External documentation fetched via wx.page and site context reports from wx.context.
Boundary markers: The skill lacks explicit boundary markers or instructions for the agent to ignore embedded commands within the ingested documentation.
Capability inventory: The skill has significant capabilities including network operations (wx.post, wx.get), file writes (fs.writeFileSync), and shell command execution (wx.bash).
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from documentation pages before it is processed by the agent.

wix-base44-connector