The Agent Skills Directory

[DYNAMIC_EXECUTION]: The lib.mjs utility script employs dynamic import() to load sibling skill modules (such as diff or deploy) from computed file system paths. This allows the skill to share hardened utilities like stealth browser sessions across different components.
[DYNAMIC_EXECUTION]: spec-capture.mjs injects locally defined functions (genericRules, groupRows, profileRules, componentLayout) into the browser context using Playwright's addInitScript. This is used to map and tag elements within the live DOM during visual capture stages.
[DYNAMIC_EXECUTION]: The skill dynamically resolves and imports the playwright library from the project's local node_modules at runtime to support its screenshot and analysis capabilities.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits a significant attack surface for indirect prompt injection as it fetches and parses raw HTML, metadata, and JSON-LD from external, untrusted websites.
Ingestion points: External HTML content is fetched by spec-fetch.mjs and parsed by spec-parse.mjs.
Boundary markers: The skill does not implement explicit boundary markers or "ignore instructions" warnings when storing scraped text in its knowledge base.
Capability inventory: The skill possesses capabilities for network fetches, file system writes, and browser automation via Playwright.
Sanitization: No sanitization of scraped text content is performed before it is consolidated into the knowledge/ output files.
[EXTERNAL_DOWNLOADS]: The skill fetches configuration and historical data from established external sources:
Fetches historical web captures from the Internet Archive (web.archive.org).
Downloads Adobe Experience Manager telemetry bundles from bundles.aem.page using an environment-provided domain key.
Retrieves tag manager configurations from Adobe's official assets domain (assets.adobedtm.com).
Fetches consent management rules from OneTrust's well-known CDN (cdn.cookielaw.org).

spec