Subagent-Driven Development (Spec Kit Plus)

What problem does it solve?

It prevents wasteful or unsafe “leader-inline” execution by ensuring subagents only run once the correct owning sp-* workflow is selected and the work is packaged with clear boundaries, evidence, and verification.

Core Features & Use Cases

• Routes to the correct owning workflow first: directs execution into canonical sp-tasks, sp-implement, sp-quick, or sp-debug before dispatching any delegated work.
• Enforces packetized execution contracts: requires a validated WorkerTaskPacket (or equivalent) with write sets, forbidden drift, acceptance checks, and verification commands before dispatch.
• Dispatches native subagents safely: chooses one-subagent for single lanes, parallel-subagents for independent lanes, and uses leader-inline-fallback only with recorded reasons when delegation is unavailable.
• Requires structured handoffs and evidence: ensures each subagent returns changed-file summaries, verification results, failures, open risks, and spec/plan gaps so work can be reviewed in order.
• Reduces routing and evidence mistakes: discourages stale strategy labels, idle “silent completion,” and code-quality review before spec compliance review.

Quick Start

Route execution to the owning workflow, build a validated WorkerTaskPacket for the lane, then dispatch native subagents from that packet while requiring structured handoffs with verification evidence.