The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill enables the agent to interact with and process data from untrusted external websites via playwright-cli, creating a potential surface for instructions embedded in web content to influence the agent.
Ingestion points: Processing of external web content through playwright-cli snapshot, playwright-cli open, and playwright-cli find commands.
Boundary markers: The instructions explicitly state that the agent should "Treat generated code and locators as observations, not trusted final tests" and "Observe before encoding."
Capability inventory: The skill utilizes Playwright to execute browser-side JavaScript, perform file system operations (saving traces and snapshots), and interact with network resources via the browser.
Sanitization: Safety is enforced through instructions requiring the agent to "Name the question or claim" before investigation and to perform human-in-the-loop review of all resulting code and locators.
[COMMAND_EXECUTION]: The skill instructs the agent to execute standard development and testing commands, such as npx playwright test, pytest, dotnet test, mvn, and gradlew. These are appropriate for the skill's stated purpose of software testing and automation.

playwright-testing