The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external websites intended for migration.
Ingestion points: Phase 1 (Extract) in SKILL.md and data processing in json-query.mjs and html-slice.mjs ingest structure and content from live URLs.
Boundary markers: Utilizes Markdown structural elements and JSON table formatting to separate data.
Capability inventory: The skill has broad capabilities including network requests via Playwright, file system writes, shell script execution (gate.sh), and deployment to Adobe's AEM production environment.
Sanitization: html-slice.mjs performs aggressive cleaning of captured HTML by removing <script>, <style>, and other non-structural elements.
[COMMAND_EXECUTION]: Orchestration of the migration phases involves executing local scripts and system commands.
Evidence: gate.sh coordinates various diagnostic tools; run-bg.mjs and run-capped.mjs utilize child_process.spawn to manage concurrent tasks and enforce operational timeouts.
[EXTERNAL_DOWNLOADS]: The skill manages the installation of necessary development and comparison tools.
Evidence: SKILL.md specifies the installation of playwright, pixelmatch, pngjs, and cheerio via npm if they are not already present in the environment.

replica