The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes content from external web pages, which presents a surface for indirect prompt injection. Ingestion points: URL navigation performed in snapshot.py, screenshot.py, verify.py, and interact.py. Boundary markers: Explicit warning messages are appended to script outputs to instruct the agent to ignore directives found within captured content. Capability inventory: The skill includes with_server.py for executing subprocess commands and interact.py for automated browser interactions. Sanitization: Content is retrieved via Playwright's synchronization API and presented as accessibility snapshots or logs.
[COMMAND_EXECUTION]: The with_server.py script manages local development server lifecycles by executing commands provided via the --cmd flag through subprocess.Popen. This is a primary function of the skill for developer workflows.
[EXTERNAL_DOWNLOADS]: Documentation in SKILL.md guides the installation of the playwright library and browser binaries from official package registries.

testing-browser