The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill enables an agent to navigate to and interact with arbitrary external websites and API endpoints using Playwright's browser automation (page.goto) and various network tools (WebFetch, WebSearch). This creates a vulnerability surface where instructions embedded in the content of an external site or API response could attempt to influence the agent's subsequent actions.
Ingestion points: External content accessed via page.goto (in REFERENCE.md), WebFetch, and WebSearch (in SKILL.md).
Boundary markers: No explicit markers are used to isolate or ignore instructions potentially found within the web pages being tested.
Capability inventory: The agent has access to shell execution (Bash), file modification (Write, Edit, TodoWrite), and directory traversal (Glob, Grep).
Sanitization: No mechanisms for sanitizing or filtering untrusted content from the browser context are mentioned.
[COMMAND_EXECUTION]: The skill provides instructions for executing various local commands via bunx and playwright. These include test execution, browser installation, and report generation. These are standard developer tools and used within the intended scope of the skill.
[EXTERNAL_DOWNLOADS]: The installation steps (bunx playwright install) fetch browser binaries and system dependencies from official Microsoft repositories. These are well-known and expected sources for the Playwright framework.

playwright-testing