Three community frameworks for AI coding agents are compared: Superpowers (enforces TDD with a 7-phase workflow and subagent delegation), GSD (prevents context rot via per-phase orchestrators that write state to disk), and GSTACK (models a 23-person team with role-based governance). Each addresses a distinct failure mode: code breaking without tests, quality degrading as context windows fill, and scope creep from unchecked agent autonomy. The post maps each framework to specific use cases — Superpowers for solo devs needing test discipline, GSD for multi-day complex projects, GSTACK for founder-engineers shipping products — and discusses how they complement Pulumi infrastructure workflows.
Table of contents
The problem all three frameworks solveSuperpowers: the test-driven discipline enforcerGSD: preventing context rot before it ruins your projectGSTACK: when you need a whole team, not just an engineerWhere each framework fitsCombining frameworks with Pulumi workflowsGetting startedSort: