The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to interact with the host system using the gbrain CLI utility for several critical operations.
Evidence: The agent is instructed to use commands such as gbrain schema show --json, gbrain capture --file <analysis.md> --slug <taxonomist-path>, gbrain sync, gbrain doctor --json, and gbrain check-resolvable throughout its operational phases.
[INDIRECT_PROMPT_INJECTION]: The skill implements an 'Always-On Auto-Fire Gate' that triggers when the agent processes user-provided documents, creating a surface where untrusted data is converted into permanent knowledge base entries.
Ingestion points: The skill logic in SKILL.md explicitly processes PDFs, file attachments, external links, and pasted long-form content.
Boundary markers: Absent. The skill does not define delimiters or specific 'ignore embedded instructions' warnings for the document content during analysis.
Capability inventory: The agent has access to put_page and add_link tools, as well as the gbrain CLI for writing to permanent storage.
Sanitization: Absent. There is no mention of validation, escaping, or filtering of external document content before it is processed and filed.

eiirp