The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted data from multiple sources to define its testing logic, creating a surface for potential prompt injection attacks.\n
Ingestion points: User-provided test steps via the $ARGUMENTS variable and project-specific files read from the tests/ and src/ directories.\n
Boundary markers: The instructions do not implement explicit delimiters or safety warnings to ensure the agent distinguishes between instructions and processed data.\n
Capability inventory: The skill is capable of modifying application source code (e.g., adding data-testid attributes), writing new executable scripts, and performing shell operations.\n
Sanitization: There is no logic provided to sanitize or validate the content of the test steps before they are used to generate code or influence agent behavior.\n- [COMMAND_EXECUTION]: The skill relies on several shell commands to fulfill its testing workflow. This includes using npx playwright test for execution, npx playwright trace for debugging, find for file discovery, and git for version control operations. While these tools are standard in development environments, their use by an agent to execute dynamically generated code represents a broad capability that should be monitored.

pw-new-test