The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from diverse and potentially untrusted sources such as emails, meeting transcripts, conversation logs, and shared links to enrich a central "brain" knowledge base.
Ingestion points: Data enters the context through Phase 2: On Every Inbound Signal, which specifically monitors "Every message, meeting, email, or conversation."
Boundary markers: The instructions do not specify any boundary markers (e.g., XML tags or delimiters) or instructions for the agent to ignore embedded commands when writing data to the brain.
Capability inventory: The skill possesses significant mutating capabilities, including put_page, add_link, add_timeline_entry, and sync_brain across multiple directories (people/, companies/, deals/, etc.).
Sanitization: There is no evidence of sanitization, validation, or filtering of the inbound signals before they are written to the permanent knowledge base.
[COMMAND_EXECUTION]: The instructions reference the use of gbrain CLI tools (e.g., gbrain entity, gbrain think, gbrain sources list) to interact with the knowledge base. While these appear to be the primary interface provided by the platform for this skill's purpose, they represent a surface for shell command execution if user input were to be improperly escaped when passed to these tools.

brain-ops