The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill interacts with and extracts data from external websites using smooth run and smooth extract, creating a significant attack surface for indirect prompt injection where malicious instructions embedded in web pages could influence the agent's behavior.
Ingestion points: Untrusted content from websites navigated to via smooth run and smooth extract (found in SKILL.md).
Boundary markers: The skill instructions do not specify the use of delimiters or specific safety instructions to isolate external data from the agent's core instructions.
Capability inventory: The skill enables broad system and network interactions, including arbitrary browser automation (smooth run), local file uploads (smooth upload-file), JavaScript evaluation (smooth evaluate-js), and credential configuration (smooth config).
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from the web before it is processed by the agent.
[DYNAMIC_EXECUTION]: The skill includes the smooth evaluate-js command, which allows the execution of arbitrary JavaScript code within the browser session at runtime. While standard for web automation, this is a form of dynamic execution that can be targeted by attackers.
[COMMAND_EXECUTION]: The skill relies on the execution of multiple CLI commands that interact with the local filesystem and shell, such as smooth upload-file for reading local documents and smooth config for managing environment variables and credentials.

smooth-browser