Skip to content

Agent profiles

Mastery HQ keeps a lightweight local profile per agent. Every task an agent runs is recorded with an inferred category (frontend, tests, infra, …) and, at the end of the run, whether it succeeded — so over time each agent accumulates a track record.

  • Work history — categorized entries per task, stamped with the outcome.
  • Skills — what this agent is good at, feeding routing decisions.
  • Backlog & cache — per-agent working state the orchestrator can consult.

When you give Hermes a goal, its planner sees a roster built from these profiles — Alex (Claude) — frontend ×12; tests ×5 — and assigns tasks to agents by name. Reliability counts, not just volume: outcomes are written back to the profile when runs finish.

  • Profiles are recorded automatically; there’s nothing to configure.
  • A dedicated profiles pane (inspecting and editing the record in-app) is planned; today the record works behind the scenes.