The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands to discover and start development servers based on project files (e.g., npm run dev, docker compose up, python manage.py runserver).
[EXTERNAL_DOWNLOADS]: The skill performs runtime installation of the agent-browser package from the npm registry if it is not already present in the environment.
[DATA_EXFILTRATION]: The skill captures screenshots and video recordings of the testing session and uploads them to external S3 storage using presigned URLs obtained via an MCP tool.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) as it processes untrusted external data that could influence its behavior.
Ingestion points: git diff output, package.json, and other project configuration files analyzed during the discovery phase.
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious instructions embedded within the files being analyzed.
Capability inventory: The skill possesses the ability to execute shell commands, automate browser interactions, and perform network uploads.
Sanitization: No sanitization or validation of the content read from the repository is performed before the agent acts upon the gathered context.

alan-test-feature