Role policy
Role and subagent machinery is configuration and spawn parameters, not a fixed pipeline. Intra-harness role policy chooses which model, prompt, and tool surface fits a subagent or specialized pass. Veyyon is provider-agnostic: roles are not hard-coded provider assumptions.
What exists today
- Subagents via the
tasktool (packages/coding-agent/src/task/executor.ts)./agentsopens the live hub for active and persisted agent threads. - Explicit model policies, not a role-to-model matrix: the interactive model (
/model), profile-wide subagent defaults and per-agent overrides undersubagent, pluscompaction.model.defaultis not a model or a role. Named roles (modelRoles, scoped per profile) let you pin specific work types. Edit roles under Settings → Model → Roles and subagent policy under Settings → Subagents. See Compaction & project memory and Models, roles, and profiles. - Plan / goal modes alter prompts and tool gating (
/plan,/goal). The advisor watchdog (advisor.enabledand related settings, inpackages/coding-agent/src/advisor/) is a background continuous-review mechanism rather than a mode you invoke;/advisorreports and configures it. Seedocs/handbook/src/features/advisor.md. - Addressed inter-agent messaging via the
irctool (packages/coding-agent/src/tools/irc.ts,packages/coding-agent/src/irc/bus.ts):send/wait/inbox/listops over a process-global bus.sendis fire-and-forget with delivery receipts; the bus wakes an idle recipient with a real turn, revives a parked one, or injects a non-interrupting aside into a busy one, the shipped analogue of wake-now-vs-defer message routing.wait(orsend await:true) observes the recipient’s reply as a real turn. Gated byisIrcEnabled: available to every subagent and to a top-level session that can still spawn subagents.
No fixed role pipeline
Veyyon does not enforce a staged plan → implement → verify → repair handoff. It uses lighter-weight
spawn, subagent-policy, and irc messaging patterns instead; you compose the stages yourself.
Pair role choice with execution-order prompts: explore → plan → edit → verify.