The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses dynamic imports and file URL resolution to load logic from neighboring skill directories within the stardust toolset.
Evidence:
scripts/checks/chrome.mjs imports chrome-compare.mjs using await import(pathToFileURL(...).href).
scripts/checks/dynamics.mjs imports dynamics-check.mjs from a relative path outside the skill directory.
scripts/from-knowledge.mjs dynamically imports eds-path.mjs from a neighboring directory.
scripts/lib.mjs uses createRequire and req.resolve to dynamically load the playwright module from the project's current working directory.
[INDIRECT_PROMPT_INJECTION]: The skill crawls and processes content from user-provided live URLs, creating an attack surface where malicious website content could attempt to influence the agent's behavior.
Ingestion points: Website content is ingested via fetchPage in routing.mjs, content.mjs, links.mjs, and metadata.mjs, and via Playwright in browse.mjs, perf.mjs, editability.mjs, and ai-readability.mjs.
Boundary markers: Absent. Findings are written to report files without explicit warnings about embedded instructions.
Capability inventory: The skill can write files to the local stardust/qa/ directory using writeJSON and execute internal scripts via spawnSync.
Sanitization: Raw data is written to report.json without sanitization; report-html.mjs performs basic HTML entity escaping for display.
[EXTERNAL_DOWNLOADS]: The skill downloads an external accessibility testing library from a well-known CDN service.
Evidence: scripts/checks/browse.mjs defines AXE_CDN as https://cdn.jsdelivr.net/npm/axe-core@4.10.2/axe.min.js and injects it into the browser context at runtime.
[COMMAND_EXECUTION]: The skill executes internal scripts using the system shell to perform specific QA tasks.
Evidence: scripts/checks/chrome.mjs uses spawnSync(process.execPath, ...) to run the chrome-explore.mjs script.

qa