pstack: per-role model configuration via /setup-pstack

[poteto]

Makes pstack's model choices configurable per role instead of hardcoded, so the plugin fits each user's model access rather than assuming a fixed set.

• New /setup-pstack skill. Detects your available models and writes ~/.cursor/rules/pstack-models.mdc, an always-applied rule with one line per role. Re-runnable and idempotent.
• poteto-mode gains a Model configuration section with the default table and precedence. Every skill reads its configured model and falls back to the inline default when the rule is absent, so it works with zero setup.
• Review panels (arena, architect, interrogate, how critics) are config-driven and count-agnostic. One subagent runs per configured model, with no fixed count or A/B/C/D labels.
• Dropped gpt-5.3-codex-high-fast. bug-fix and perf-issue default to gpt-5.5-high-fast; feature and refactoring stay composer-2.5-fast.
• Version bump to 0.8.0.

---

Note

Low Risk
Changes are limited to plugin docs/skills and a user-local Cursor rule template; no application runtime or shared infra behavior changes.

Overview
Adds /setup-pstack so users can map pstack roles (feature work, judgment, arena/architect/interrogate panels, etc.) to models they actually have access to. The skill detects available Task model slugs, confirms choices, and writes ~/.cursor/rules/pstack-models.mdc as an always-applied override; skills keep inline defaults when a line is missing.

Across poteto-mode, playbooks, and workflow skills (how, why, arena, architect, interrogate, reflect), hardcoded model tables are replaced with configured role + default fallback wording. Multi-model review flows are count-agnostic (one subagent per configured model, no fixed A/B/C/D roster). gpt-5.3-codex-high-fast is dropped from defaults; bug-fix and perf-issue default to gpt-5.5-high-fast. Plugin version 0.8.0 and README document the new skill.

^{Reviewed by Cursor Bugbot for commit be280e7. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{Bugbot Autofix is ON, but it could not run because the branch was deleted or merged before autofix could start.}

^{Reviewed by Cursor Bugbot for commit df0b177. Configure here.}

[cursor]

Architect arena runner mismatch

Medium Severity · Logic Bug

Phase B tells agents to use configured architect runners, but it invokes arena, whose Frame step picks configured arena runners instead. After /setup-pstack, those roles can differ, so design sketches may fan out with the wrong models despite the architect override.

Additional Locations (1)

• pstack/skills/arena/SKILL.md#L27-L28

 

^{Reviewed by Cursor Bugbot for commit df0b177. Configure here.}

[cursor]

Empty panel disables review

Medium Severity · Logic Bug

Panel roles now spawn one subagent per configured model, but setup-pstack never requires a non-empty list and skills only fall back when a role line is absent. A present line with no models yields zero reviewers or critics, so flows like interrogate can complete with no adversarial input.

Additional Locations (1)

• pstack/skills/interrogate/SKILL.md#L35-L36

 

^{Reviewed by Cursor Bugbot for commit df0b177. Configure here.}