Configure per-role model assignments for pstack skills in an always-loaded override file.
npx skills add https://github.com/adjohn/pstack --skill setup-pstack-adjohn
Please help me install this Agent Skill. Skill: setup-pstack Source: https://github.com/adjohn/pstack/tree/main/skills/setup-pstack Command: npx skills add https://github.com/adjohn/pstack --skill setup-pstack-adjohn
~/.claude/pstack-models.md with one line per role and imports it into every session via ~/.claude/CLAUDE.md. - Validation and Idempotency: Rejects unavailable slugs, supports inherit-parent and auto aliases, and re-runs cleanly by overwriting the whole file. - Use Case: After installing pstack, run the setup to map roles like critics, arena runners, and swarm workers to the Claude tiers your account exposes, so multi-agent workflows never fail on an unavailable model. ## Quick Start Ask the assistant to run /setup-pstack to detect your available models and configure pstack's model assignments.High-intent search queries and answers about installing and using this skill.
Run the /setup-pstack skill, which detects your available Claude models and writes ~/.claude/pstack-models.md with one role-to-model line per entry. The file is imported into every session via ~/.claude/CLAUDE.md, and re-running the skill updates it.
There is no in-session API listing entitled models, so the skill either test-spawns a one-line background subagent per candidate slug and treats a rejection error as unavailable, or asks you directly which models your account exposes.
Yes. Setting a role's value to inherit-parent or auto makes that role run on the parent chat model by omitting the Agent model parameter. These aliases are always valid even though they are not detected slugs.
An override pointing at an unavailable model breaks every delegation that reads it, so the skill validates every real slug against the detected set before writing. If a chosen slug is unavailable, it stops and asks you to choose again.
Panel roles like critics, arena runners, and interrogate reviewers accept a comma-separated list of models, and one subagent runs per entry, so the list length sets the fan-out count. Alias entries in the list still count toward the total.
hmctsDiscover, install, and manage Claude Code plugins from the Agentic Plugins Marketplace.
cursorConfigure per-role model assignments for pstack by writing an always-applied Cursor rule.
cursorGenerates a project-local verification skill that drives an app like a user and captures evidence.
orenmagidRead the skill index and group discoverable skills into tables.