Subagents and the research panel
What agent-brand-discovery and agent-brand-structuring do. What the operator sees.
Brand Crafter does not work alone. Two specialist subagents handle research and structuring under the supervisor's direction.
The two subagents
agent-brand-discovery
Brand intake specialist. Two modes:
MODE: pre_research. A fast public-only scan before the supervisor's first strategic question. About 30 to 60 seconds.MODE: full_discovery. A thorough scan, run when the operator has shared documents or when a complete knowledge graph is required.
Returns: a structured fact sheet and the brand foundation fields with provenance.
agent-brand-structuring
Brand memory author. Called only after validation. The supervisor must hold every required field with a status and a source before this subagent runs.
Returns: machine-readable brand memory, decision objects, guardrail records, and confidence scores.
How the operator experiences the subagents
The supervisor is the only voice the operator talks to. The subagents do their work underneath. The interface shows the difference:
- The chat pane stays in the consultant register. Brand Crafter speaks.
- Research panels expand below or beside the chat. The subagents' working notes show there.
- Tool calls render as small status rows: started, finished, errored.
The operator does not need to read the research panels to follow the conversation. They are there when a claim needs checking.
What the subagents will not do
- Write to Brand Brain. Only the supervisor publishes.
- Speak directly to the operator. Their output is source material for the supervisor.
- Reveal their internal names. The chat does not say "agent-brand-discovery found three competitors." The chat says "I found three competitors."