Company Brain / Project skills
Brain Setup
- name
- brain-setup
- description
- Use when the tables described in /brain/SETUP.md do not exist yet, the user asks to set up the knowledge base or company brain, or list_tables shows a table from the setup doc is missing.
Goal
Create the machinery the brain needs, and leave behind a taxonomy and an intake document that match this company rather than the template's guesses.
Steps
- Call list_tables. If all three tables already exist, stop and say so — do not recreate a table a human may have adjusted on purpose.
- Read /brain/SETUP.md. It is the specification; you are executing it. If a column looks wrong, raise it before creating anything.
- Ask what the company does, what areas knowledge falls into, and who owns each area. Three questions. Do not interview the user about columns.
- Confirm the plan in a few lines, then create the three tables with create_data_table exactly as specified.
- Create /brain/inbox/README.md and /brain/archive/README.md with short headers saying what belongs in each and what happens to it.
- Rewrite /brain/TAXONOMY.md with the company's real areas and owners. Delete the starter set — leaving both is how a taxonomy becomes ambiguous on day one.
- Rewrite /brain/INTAKE.md, replacing every placeholder with the real table id and column ids read back from list_tables.
- Write the company context into /SUBTERRANEAN.md.
- Report what was created, then give the activation step verbatim: open a session with Curator and say "activate".
Do not
- Do not skip steps 6 and 7. A brain with a generic taxonomy and a placeholder intake document is a folder, not a brain.
- Do not write knowledge pages during setup. Setup builds the shelves; Archivist fills them.
- Do not configure another agent's triggers or heartbeat. That is step 9's job, in Curator's own session.