plan-large

To plan a large feature as incremental pull requests, you provide a feature description to generate a DAG of smaller increments. This produces a structured plan document specifying explicit dependencies, increment ordering, done criteria, and file specs for implementation.

A DAG-based increment plan models a large feature as a sequence of smaller, ordered increments with explicit dependencies. You need it for architecture review when a feature spans multiple pull requests and requires upfront decisions to be traceable across the implementation pipeline.

You generate a structured plan document for multi-PR implementation by providing a feature description, which creates a .plans/<slug>/plan.md file. This file captures architectural decisions, increment ordering, done criteria, and file specifications for each stage of the work.

Yes, the plan generation process includes a code-search-first gate followed by a blind self-review pass. These multi-pass reviews critique and improve the generated plan before presenting it, ensuring the increment ordering and file specifications are validated against the existing codebase.

The best way to break down large architecture changes is modeling the work as a directed acyclic graph where each node is a small increment. This captures architectural decisions upfront and ensures explicit dependencies, done criteria, and file specs are documented for traceable execution.

Yes, you can use this incremental planning approach for features requiring strict PR ordering because it models increments as a DAG. This ensures explicit dependencies between increments are documented, allowing teams to execute pull requests in the correct sequence.