subagent-dispatch

Protocol for handing work to subagents or to prompts the user runs one per chat: the model per role, a brief that carries the task instead of the session history, artifacts as files, a task ledger each run checks before building on another's, the four report statuses, and a capped fix loop. Load it whenever work is split or delegated, even with no subagent. USE FOR: dispatch a subagent, which model for a subagent, verify a subagent's work, the agent said it was done, re-review, fix loop, blind re-performance, resume after compaction, handoff prompts for separate chats, run the next brief, task ledger, migrate a prompt folder. DO NOT USE FOR: the review rubric itself (use code-review-and-quality), authoring a Custom agent or Skill (skill-creator), measuring a Customization (agent-evals), monitoring a job (long-running-job-monitor), a single handoff document (session-handoff prompt), durable project knowledge (memory-bank), unattended spec completion (spec-completion-controller).

Compatibility

Requires a harness that can dispatch subagents and select their model explicitly, such as the VS Code Copilot chat agent surface or the GitHub Copilot CLI. Ledger and artifact handover assume a writable working tree. The ledger scripts need PowerShell 5.1 or later, and checking a row's commits needs git.