The Agent Skills Directory

[COMMAND_EXECUTION]: The bootstrap.md file contains instructions to execute npx -y skills add via child_process.execSync with shell access (/bin/bash). This pattern allows for arbitrary command injection if the source string or environment variables are manipulated.
[REMOTE_CODE_EXECUTION]: The SKILL.md file instructs the agent to execute a locally downloaded JavaScript file (scripts/docs.js) using the new Function constructor (an eval-like mechanism). This bypasses standard module loading safety and executes arbitrary code strings at runtime.
[EXTERNAL_DOWNLOADS]: The skill initiates multiple network operations to Wix-related domains (www.wixapis.com, mcp.wix.com, dev.wix.com) to fetch documentation and API specifications. While these are vendor-owned, the use of bootstrap.md to pull code from a remote registry adds a layer of supply-chain risk.
[CREDENTIALS_UNSAFE]: The callApi function in scripts/docs.js is designed to accept and use Authorization: Bearer tokens. Instructions in SKILL.md explicitly direct the agent to retrieve accessToken from the base44 connector and pass it to this tool, creating a path for credential handling and potential exposure if the URL targets are redirected.

wix-docs-base44