The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill mirrors entire websites from external domains using the wget utility. It is configured to span hosts and download assets from common website-builder CDNs such as cdn.prod.website-files.com, framerusercontent.com, and static1.squarespace.com.
[COMMAND_EXECUTION]: The migration process involves a high volume of shell command executions, including wget for mirroring, sed for rewriting URLs in bulk, and grep for extracting asset links. It also uses embedded Python one-liners for tasks like extracting webpack chunk hashes from JS bundles and validating asset references across all HTML files.
[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection because it downloads and then processes untrusted HTML, CSS, and JavaScript content from the internet.
Ingestion points: The wget mirror operation downloads raw content into the local filesystem (e.g., SKILL.md refers to the site/ directory and mirrored domain data).
Boundary markers: None are present; the skill treats the downloaded files as data to be parsed and modified without explicit isolation or 'ignore instructions' warnings.
Capability inventory: Across the workflow and the mirror-site.sh script, the skill can read/write to the filesystem, execute shell commands, run Python scripts, and perform network operations for deployment (e.g., via Netlify or Vercel CLIs).
Sanitization: The scripts use regular expressions to extract specific data but do not sanitize the source content before it enters the processing pipeline.

website-builder-migrate