Skip to content
Back to skills

Setup Pstack

ASecurity

Configure which models pstack uses per role. Detects the models your Agent tool accepts and writes `~/.claude/pstack-models.md`, which overrides the skill defaults. Use for /setup-pstack, "configure pstack models", or changing pstack's model choices.

  • 2 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added October 8, 2026
ai-agentsbashrefactoringapi

Works with

  • claude code
  • cli
  • api

Security analysis

A100/100

Scanned October 8, 2026

npx -y skills add chillerno1/kskills --skill setup-pstack --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Setup Pstack?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Setup Pstack
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/chillerno1-setup-pstack/badge)](https://www.skillsdirectory.com/skills/chillerno1-setup-pstack)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: setup-pstack
description: Configure which models pstack uses per role. Detects the models your Agent tool accepts and writes `~/.claude/pstack-models.md`, which overrides the skill defaults. Use for /setup-pstack, "configure pstack models", or changing pstack's model choices.
---

# Setup pstack

Write `~/.claude/pstack-models.md`, a plain markdown file that sets pstack's model per role. When the user asks for a per-repo override, write `.claude/pstack-models.md` at the repo root instead. A project file wins over the user file. If the target file has `## ` sections (pstack's shipped `pstack-models.md`, usually symlinked), this skill reads and rewrites only the `## Claude Code` section's role lines and leaves the header and the Grok and Codex sections as they are.

Reasoning effort is not part of this file. Claude Code cannot set effort per Agent call. It comes from the agent definition, and `poteto-agent` pins `effort: high` in its frontmatter. To change effort for code work, edit that one line in `agents/poteto-agent.md`. Explorers and investigators that run as `Explore` or `general-purpose` follow the session's effort.

## Steps

### 1. Detect available models

Read the `model` parameter's enum on your `Agent` tool. That enum is the detected set. Do not call any API or CLI to list models. Today it is `opus`, `fable`, `sonnet`, `haiku`. The value `inherit` is always valid even though it is not in the enum. It means the role runs on the parent session model. Never write a model you have not confirmed is in the enum.

### 2. Load current state

The default role-to-model mapping is the file shape shown in step 5 below. If the target file already exists, read it and treat its role values as the current choices. Otherwise start from those defaults. A line whose role is not in step 5, such as `how critics`, is from a retired role. Drop it. A `# budget` line is retired too. Drop it. A value of `inherit-parent` or `auto` is the old spelling of `inherit`. Rewrite it as `inherit`.

### 3. Show the roles and confirm

Build the working table from the skill defaults, and on a re-run keep any role whose value differs from the default (a different model, a different list, or `inherit`). Show every role with its model as a table in your reply, marking any value not in the detected set as needing a choice. Also list each line step 2 dropped.

Then use AskUserQuestion: accept as-is, or change specific roles. For each role the user wants to change, ask again with the detected models as the options. `inherit` and panel lists come in through the free-text answer. For panel roles (arena runners, architect runners, interrogate reviewers) the value is a list, and one subagent runs per entry, `inherit` entries included, so the list length sets the count. `arena cross-judge pool` is also a list, but Arena selects one value from it that differs from the parent session's model when possible. `swarm workers` is the default model for every worker unless a race or comparison assigns another model per arm.

Every value is a Claude model, so panels no longer mix model families. Their diversity comes from different Claude models plus distinct reviewer personas and prompts. If `codex` is on PATH, a panel may add one `codex exec` run via Bash as an optional non-Claude seat. It is never required and never written to this file.

### 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.

### 5. Write the file

Write the target file with no frontmatter and one line per role, using the same labels poteto-mode uses. Overwrite the whole file so re-runs stay idempotent. Shape:

```
# pstack model configuration. One line per role. Delete a line to fall back to the skill default.
# Values: opus, fable, sonnet, haiku, or inherit (the role runs on the parent session model). inherit entries in a panel list still count toward its fan-out.
# Skills read this file on demand. It is not auto-loaded. A repo's .claude/pstack-models.md wins over ~/.claude/pstack-models.md.
# Effort is not set here. poteto-agent pins effort: high in its frontmatter.
feature, refactoring: opus
bug-fix: opus
perf-issue: opus
hillclimb: opus
judgment and prose: opus
hardest tasks: opus
how explorer: opus
how explainer: opus
why investigators: opus
why synthesizer: opus
reflect tooling: fable
reflect judgment, divergent, synthesizer: opus
arena runners: fable, opus
arena cross-judge pool: fable, opus, sonnet
swarm workers: opus
architect runners: fable, opus
interrogate reviewers: fable, opus, sonnet
```

### 6. Confirm

Tell the user where the file was written. Say that pstack skills read it on demand each time they route a model, and that it is not auto-loaded into context, so it takes effect on the next skill run, this session included. 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` (resolves wherever pstack is installed: project, user, or plugin). On no, move on without pushing.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.