The Agent Skills Directory

[COMMAND_EXECUTION]: The skill leverages the qa-studio command-line interface to manage tests. Actions include listing tests (qa-studio tests list), importing test definitions (qa-studio tests import), and executing test runs (qa-studio run). These operations are essential to the skill's primary function of test automation.
[DATA_EXFILTRATION]: When running tests, the system generates artifacts such as screen recordings, screenshots, and execution logs. By default, these are uploaded to the vendor's cloud infrastructure (AWS) to provide a reporting interface, though a --local-only mode is available for users who wish to keep data on-disk only. The skill also handles sensitive login credentials using a 'secret' mechanism that prevents values from being stored in test scripts or logs.
[EXTERNAL_DOWNLOADS]: The documentation references the installation of the qa-studio package and its optional agentcore component via standard package managers like pip. These resources originate from the vendor's official repositories.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it processes untrusted content from web pages and network APIs during test execution. It implements several defensive measures to mitigate this risk:
Ingestion points: Data enters the system via page extraction steps (retrieve_value, validation) and network traffic interception (network_assertion).
Boundary markers: Instructions are encapsulated within structured JSON objects, separating logic from data.
Capability inventory: The agent can execute the qa-studio CLI, write test configuration files, and interact with websites through a controlled browser environment.
Sanitization: The network_assertion step explicitly rejects external $ref URIs in JSON schemas to prevent SSRF attacks, and the transform operation restricts string formatting to simple positional placeholders to prevent template injection.

qa-studio