Agency

Agency

On-demand expert consultation. 19 specialized AI personas you invoke explicitly — never auto-delegated.

Dispatch

Parse $ARGUMENTS as follows:

• If the first word is list → Read commands/list.md and follow its instructions.
• If the first word is find → Read commands/find.md and follow its instructions. The remaining text after find is the task description.
• If the first word is consult → Read commands/consult.md and follow its instructions. The remaining text after consult contains @agent names and a task description.
• If the first word is deliberate → Read commands/deliberate.md and follow its instructions. The remaining text after deliberate contains @agent names and a task description.
• If the first word is close to one of the above (e.g., cosult, diliberate, lst, fnd) → suggest the correct command: "Did you mean /agency {correct}?" and show usage.
• Otherwise → show usage:

Usage:
  /agency list                                 — browse all agents
  /agency find [task description]              — get agent recommendations for a task
  /agency consult @agent [@agent ...] task             — get expert perspectives (1+ agents)
  /agency deliberate @agent @agent [@agent ...] task  — structured multi-expert debate (2+ agents)

Conventions

• Agent names use @ prefix: @software-architect, @planner, @readiness-advisor
• The @name is the filename stem — the agent file is at agents/{name}.md
• If an @name doesn't match a file, suggest close matches from the agents/ directory