The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill dynamically generates test instructions by processing external ticket summaries and git diffs, which could potentially introduce malicious commands into the automated browser walkthrough loop.\n
Ingestion points: Processes git diff, ticket_url, task_summary, and implementation_summary from external ticket sources to drive the browser walk.\n
Boundary markers: The instructions do not specify any delimiters or sanitization logic to isolate external data from the prompt instructions.\n
Capability inventory: Possesses the ability to execute shell commands (lt, pnpm, curl) and perform full browser automation (navigation, interaction, and JS execution) via MCP.\n
Sanitization: No sanitization of the input ticket or diff content is performed before generating the test plan.\n- [COMMAND_EXECUTION]: The skill executes various local CLI tools to manage the development environment and perform smoke tests.\n
Evidence: Utilizes lt dev up, lt dev status, lt dev init, pnpm run seed, and curl for stack management and API verification.\n- [DYNAMIC_EXECUTION]: The skill uses runtime script injection in the browser to simulate edge cases and failures.\n
Evidence: Employs the evaluate_script tool within the Chrome DevTools MCP to mock 500 errors, mutate requests, or force specific application states during validation.

validating-changes-in-browser