The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the playwright package and browser binaries (Chromium) from the official NPM registry and Microsoft's distribution servers if they are not already present in the environment. These are well-known, trusted sources.
[COMMAND_EXECUTION]: The skill executes several commands to perform its primary function, including:
Node.js scripts for image processing and manifest management (shoot.mjs, finalize.mjs, png-check.mjs).
Mobile development tools (adb for Android, xcrun simctl for iOS) to capture native application screens.
NPM for dependency management.
Playwright for browser automation.
[INDIRECT_PROMPT_INJECTION]: The skill captures content from external web pages, which is a potential surface for indirect prompt injection. However, the skill provides specific instructions to the agent to treat page messages, errors, and URLs as untrusted text and fences them appropriately in the output to mitigate risk.
[SAFEGUARD_MECHANISMS]: The skill implements several internal safeguards:
Path Validation: Filenames for screenshots are strictly validated against a regular expression (/\d{2}-[a-z0-9]+(?:-[a-z0-9]+)*/) to prevent path traversal attacks during file writes or deletions.
URL Enforcement: The shoot.mjs script validates that target URLs use only http: or https: protocols, preventing the tool from being used to capture sensitive local files via file:// schemes.
Data Masking: Instructions include rules for masking volatile or sensitive content (timestamps, IDs, avatars) to ensure deterministic and safe captures.

capturing-screenshots