The Agent Skills Directory

[COMMAND_EXECUTION]: The script compare-pages.mjs uses node:child_process.execFile to launch a headless Google Chrome or Chromium process. This is used to take screenshots of both the live and local versions of a page for visual regression testing.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external websites and local Webflow export files, which could potentially contain malicious instructions intended to influence the agent's behavior during the migration process.
Ingestion points: Content is retrieved via fetch in compare-pages.mjs and readFileSync in scan-export.mjs, map-classes.mjs, map-variables.mjs, and split-css.mjs.
Boundary markers: None present to distinguish untrusted external content from system instructions.
Capability inventory: The skill can execute shell commands (Chrome via execFile), read/write to the file system (screenshots and reports), and perform network operations (fetch).
Sanitization: Content is processed using regular expressions for comparison, but no formal sanitization is applied to the raw input data.
[EXTERNAL_DOWNLOADS]: compare-pages.mjs uses the fetch API to download the HTML content of the live Webflow site being migrated and the local development server to perform content parity checks.
[DYNAMIC_EXECUTION]: The scripts map-classes.mjs and map-variables.mjs include a --sed mode that generates shell scripts (e.g., rename-classes.sh) using string concatenation. These scripts are provided for the user to run manually in their terminal to automate project-wide renaming of classes and CSS variables.

lumos-import-webflow