The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages a local Node.js server lifecycle using shell scripts (start-server.sh, stop-server.sh) and a core server script (server.cjs). It correctly implements process monitoring, background execution using nohup and disown, and graceful shutdown procedures.
[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection (Category 8) by design. \n
Ingestion points: Data from the user's browser (clicks, selected choices, and form field values) is captured by helper.js, transmitted via WebSocket or HTTP to server.cjs, and stored in a local events file. This data is then consumed by the agent using wait-for-event.cjs. \n
Boundary markers: The instructions do not specify any boundary markers or instructions for the agent to treat the data retrieved from browser events as untrusted or to ignore embedded commands. \n
Capability inventory: The skill is configured with Bash, Read, and Write tools, which gives the agent significant capabilities should it be misled by injected instructions in the event data. \n
Sanitization: There is no evidence of sanitization or validation of the text or values returned from the browser-based forms before they are processed by the agent.

visual-companion