Agent Skills: Setup pstack

Setup pstack

Write ~/.claude/pstack-models.md, the config file that sets pstack's model per role. Every pstack skill reads it and falls back to its inline default when a line is absent, so this is an override layer, not a requirement.

A project can override the user-level file with .claude/pstack.local.md in the repo root, same shape. Project beats user; a role absent from both keeps the skill default. Add .claude/*.local.md to the repo's .gitignore.

Value syntax

A role's value is <model> or <model>[<effort>].

• <model> is one of opus, fable, sonnet, haiku, or inherit.
• inherit means the role runs on the parent session's model. Spawn it by omitting model on the Agent call.
• [<effort>] is optional and one of low, medium, high, xhigh, max. The Agent tool has no effort parameter and ignores it. A fan-out routed through the Workflow tool passes it as opts.effort. Carry the suffix through either way, so the same config serves both.

A panel role's value is a comma-separated list. One subagent runs per entry, inherit entries included, so the list length sets the fan-out width.

Steps

1. Detect available models

Enumerate the models you can pass to an Agent subagent in this session. The Agent tool's own schema is the dependable source; its model enum is the entitled set. Never write a model you have not confirmed is spawnable. inherit is always valid.

2. Load current state

The default role-to-model mapping is the file shape in step 5. If ~/.claude/pstack-models.md already exists, read it and treat its values as the current choices. Otherwise start from the defaults.

3. Map and confirm

Show every role with its current model, marking any value not in the detected set as needing a choice. Ask whether to accept as-is or change specific roles, offering the detected models plus inherit. Prefer AskUserQuestion over free text.

For panel roles (how critics, arena runners, architect runners, interrogate reviewers) the value is a list and its length sets the fan-out. arena cross-judge pool is also a list, but Arena selects one entry from it whose model differs from the parent's when possible. swarm workers is the default model for every worker unless a race or comparison assigns another model per arm.

Panels get their diversity from four model tiers plus per-entry effort. Two entries on the same model at different efforts are a real difference in behavior, not a duplicate, and the panel skills give each reviewer a distinct lens on top. Keep at least two distinct models in any panel.

4. Validate

Every model written must be in the detected set; inherit always passes. If a chosen model is not available, stop and ask again. A config pointing at a model the user cannot spawn breaks every delegation that reads it.

5. Write the config

Write ~/.claude/pstack-models.md with one line per role, using the same labels poteto-mode uses. Overwrite the whole file so re-runs stay idempotent. Shape:

---
description: pstack per-role model choices (overrides skill defaults)
---
# pstack model configuration. One line per role. Delete a line to fall back to the skill default.
# Value is <model> or <model>[<effort>]. `inherit` runs the role on the parent session's model (omit `model` on the Agent call).
# Effort applies only to fan-outs routed through the Workflow tool; the Agent tool ignores it.
feature, refactoring: sonnet[high]
bug-fix: opus[high]
perf-issue: opus[high]
hillclimb: opus[high]
judgment and prose: fable[max]
hardest tasks: fable[max]
how explorer: sonnet[high]
how explainer: fable[max]
how critics: fable[max], opus[high], sonnet[high], opus[max]
why investigators: sonnet[high]
why synthesizer: fable[max]
reflect tooling: opus[high]
reflect judgment, divergent, synthesizer: fable[max]
arena runners: fable[max], opus[high], sonnet[high], opus[max]
arena cross-judge pool: fable[max], opus[high], sonnet[high], opus[max]
swarm workers: sonnet[high]
architect runners: fable[max], opus[high], sonnet[high], opus[max]
interrogate reviewers: fable[max], opus[high], sonnet[high], opus[max]

The defaults split work by model strength. Fast mechanical code goes to sonnet. Precisely specified sequences and debugging go to opus at high effort. Prose and judgment go to fable. The hardest calls go to opus[max].

6. Confirm

Tell the user the config was written and that skills read it on their next run. Re-running this skill updates it.

7. Offer a verification skill (optional)

Check whether the project has a way to drive the real app for proof (a verify-* skill, or an existing harness). If not, offer once: "want a project-local verification skill, so agents can drive the app the way a user does and prove changes work? I can generate one with /create-verification-skill." On yes, invoke /create-verification-skill. On no, move on without pushing.