The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several internal Python and Node.js scripts (e.g., scripts/parity_map.py, scripts/capture-snapshots.mjs, scripts/diff_snapshots.py) to manage verification workflows, capture UI state, and perform diffing analysis.
[EXTERNAL_DOWNLOADS]: The skill requires the playwright package for browser automation. As documented in references/capture.md, the tool expects this dependency to be present in the host environment (via NODE_PATH or node_modules).
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill interacts with external web pages and retrieves DOM subtrees for comparison.
Boundary markers: Uses structured manifest and ledger files to scope the analysis, along with specific React root selectors.
Capability inventory: Includes script execution (Python/Node), file system writes (for snapshots and session ledgers), and network access via the Playwright-driven browser.
Sanitization: Extracted DOM data is parsed and normalized into a rigid JSON schema before comparison.
[DYNAMIC_EXECUTION]: The skill injects local JavaScript files (scripts/find-react-roots.browser.js and scripts/snapshot-subtree.browser.js) into the browser context of target applications using Playwright's page.addScriptTag and page.evaluate methods.

visual-parity-verification