What problem does it solve?
Pstack delegates work to subagents across many roles, and each role needs an appropriate model. This Skill detects which models you actually have access to and writes a single override rule so every pstack delegation uses the models you choose instead of the built-in defaults.
Core Features & Use Cases
- Model Detection: Enumerates the model slugs available to Task subagents in your session, falling back to asking you to paste slugs when detection fails.
- Role-to-Model Mapping: Shows every pstack role (critics, arena runners, swarm workers, interrogate reviewers, and more) with its current model and lets you confirm or change each one, including list-valued panel roles.
- Idempotent Rule Writing: Writes
~/.cursor/rules/pstack-models.mdc with alwaysApply: true, overwriting the whole file so re-runs stay consistent.
- Use Case: You want pstack's code critics to run on a reasoning-heavy model while swarm workers use a faster one. Run this Skill, pick models per role from your detected set, and the rule applies to all future sessions.
Quick Start
Ask the agent to run /setup-pstack to configure which models pstack uses for each role.