spec-pipeline

Automating the feature development pipeline requires sequentially invoking subagents for brainstorming, grill-me, writing-plans, and executing-plans to handle each phase without manual transitions. This interconnected workflow transfers only the produced document between phases, maintaining clear context.

A subagent architecture for requirement management utilizes four distinct subagents to handle different phases of the pipeline, ensuring clean context for each stage. It transfers only the produced artifact between phases, preventing context bleed across the brainstorming, grilling, planning, and executing stages.

To start the spec pipeline for a new feature, run the skill with the feature's name by executing the command `spec-pipeline <feature-name>`. This initiates the comprehensive pipeline, automatically progressing through brainstorming, grilling, planning, and executing phases.

The spec-pipeline workflow does not require any external dependencies to execute the requirement-to-implementation pipeline. It internally coordinates multiple stages and artifacts using its built-in scripts and subagent architecture.

The best way to stress-test a feature design before writing implementation plans is using the integrated grilling phase within the pipeline. It sequentially stress-tests the brainstorming output by passing only the artifact to the grill-me subagent, ensuring the design is validated before planning.

An end-to-end pipeline for feature implementation should not be used when you only need to execute an isolated phase like brainstorming or planning, rather than a complete, integrated process. It is designed for users seeking a full workflow with coordinated artifacts across all stages.