The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a Node.js script (ima_api.cjs) located in a sibling directory (../ybrain/scripts/) to handle communication with the IMA API.
[DYNAMIC_EXECUTION]: The skill invokes an external script using a path computed relative to its own installation directory, passing arguments including API paths and JSON bodies.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of stored content.
Ingestion points: Untrusted data enters the context via the search_knowledge and search_note functions which retrieve content from the "memory layer", "thought layer", and "archived volumes".
Boundary markers: No specific delimiters or safety instructions are provided to the agent to isolate the retrieved content from the skill's operational logic.
Capability inventory: The skill can execute shell commands via node and perform write operations to the notebook system using import_doc and append_doc.
Sanitization: The process relies on a manual "Feynman check" performed by the AI rather than programmatic escaping or sanitization of the input data.
[DATA_EXFILTRATION]: The skill instructions state that credentials for API requests are automatically retrieved from the sensitive file path ~/.config/ima. Accessing configuration files in the home directory that contain tokens or keys represents a potential data exposure risk.

ybrain-express