The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides the kernel.browsers.playwright.execute() method, which sends JavaScript or TypeScript code strings to be executed on a remote server. This is the primary mechanism for low-latency, server-side browser automation within the Kernel ecosystem.
[DYNAMIC_EXECUTION]: The playwright.execute method dynamically evaluates the string provided in the code parameter. This allows the agent to generate and run automation scripts at runtime.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill enables the agent to navigate to and process content from arbitrary websites using page.goto() and high-level automation libraries like Stagehand and Magnitude, as seen in examples/examples.md.
Boundary markers: The skill does not include specific boundary markers or instructions to ensure that the agent ignores commands or instructions embedded within the third-party web content it processes.
Capability inventory: The skill includes powerful capabilities such as server-side code execution, file system access via kernel.browsers.fs.readFile(), and session management including profiles and proxies.
Sanitization: There is no evidence of sanitization, validation, or filtering of external data before it is used to drive the agent's automation logic or code generation.
[COMMAND_EXECUTION]: The documentation in SKILL.md instructs the user to perform npm install commands for the @onkernel/sdk and playwright-core packages.

kernel-typescript-sdk