The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the gbrain CLI utility to perform system-level tasks, data extraction, and maintenance. This includes running background processes with nohup and utilizing the --remediate --yes flag for autonomous modification of the user's filesystem.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface by ingesting external data (transcripts and markdown files) to synthesize new content. \n
Ingestion points: Files located in the ~/brain directory and transcripts from the configured session_corpus_dir.\n
Boundary markers: The skill implements allowed_slug_prefixes and references _brain-filing-rules.json to constrain subagent write operations. \n
Capability inventory: Includes the put_page tool and various gbrain CLI commands that can create or modify files and database links.\n
Sanitization: Employs slug discipline (lowercase alphanumeric validation) and content-hash suffixes to prevent silent overwriting of existing files.\n- [PERSISTENCE]: The autopilot --install command establishes a continuous background loop (daemon) that runs sync, extract, and embed tasks, ensuring the skill's operations persist across sessions.\n- [DYNAMIC_EXECUTION]: Features a "synthesis subagent" that autonomously generates new pages and reflections based on transcript analysis, representing a form of runtime content generation and execution of LLM-derived instructions.

maintain