The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill relies on the npx neuroskill command, which downloads and executes a package from the public npm registry at runtime. This introduces an external dependency that is fetched during use.
[COMMAND_EXECUTION]: The skill uses various shell commands to interact with the NeuroSkill app, including process discovery using lsof and data visualization by piping to the dot command. User-provided input is passed as arguments to shell commands (e.g., npx neuroskill label "text"), creating a potential risk for command injection if input is not properly handled by the underlying system.
[INDIRECT_PROMPT_INJECTION]: The skill integrates biometric data and historical annotations into the agent's context, creating an attack surface for indirect instructions.
Ingestion points: Data is ingested from the local NeuroSkill server via status, search-labels, and interactive commands, specifically in the text fields of labels.
Boundary markers: The instructions provide guidelines for interpreting metrics but do not define strict delimiters or instructions to ignore malicious text embedded in retrieved labels.
Capability inventory: The skill possesses the ability to execute shell commands (npx), send OS notifications, and perform local network requests to the BCI server.
Sanitization: No explicit sanitization or filtering of biometric scores or label descriptions is performed before the agent processes the information.

neuroskill-bci