nexxeln/oc-pstack/setup-pstack
Claude Code and agent skill by nexxeln/oc-pstack
- community
Checking that the SKILL.md file is still in the repository.
Configure which models pstack uses per role and at what reasoning budget. Detects your available models and stores them with pstack_roles. Use for /setup-pstack, "configure pstack models", "pstack budget", or changing pstack's model choices.
Install
npx vskill@latest install nexxeln/oc-pstack/setup-pstackSKILL.md
Setup pstack
Store pstack's model per role with the pstack_roles tool. Every pstack skill reads its role from there. Unset roles use the newest available Opus or Grok model, preferring the session's provider. The default budget is large, with xhigh reasoning. A missing model family falls back to inherit-parent, which runs on the parent session's model.
Steps
1. Detect available models
List models with opencode.models (under code mode, tools.opencode.models(...) inside execute). Page with offset until next is null. Record each providerID/modelID and its variants. Never write a model you did not see there. inherit-parent is always valid.
2. Load current state
Call pstack_roles with no arguments. It returns every role, its current value, and the stored budget.
3. Budget, map, and confirm
(a) Ask for a budget with the question tool. Offer these labels, and name the current budget when one is stored. With no stored budget, say that large matches the skill defaults.
unlimited, max reasoninglarge, xhigh reasoningmedium, high reasoningsmall, medium reasoning
(b) Apply it. Keep any current role choices by model family, panel list, or inherit-parent. The budget picks a variant. For each chosen model, use the variant named for the target effort (max, xhigh, high, medium). If that variant does not exist, use the highest listed variant below it on max > xhigh > high > medium > low. If the model has no variants, write it without one. inherit-parent does not change. unlimited raises Opus to max when available. A Grok model that tops out at xhigh stays there.
(c) Propose roles. pstack splits work by model strength.
- Grok handles
feature, refactoring,bug-fix,perf-issue,hillclimb,how explorer,why investigators,swarm workers, andreflect toolingby default. - Opus handles
judgment and prose,hardest tasks,how explainer,why synthesizer, andreflect judgment, divergent, synthesizerby default. - Panels default to one Opus and one Grok.
arena runners,architect runners, andinterrogate reviewersspawn one subagent per entry.arena cross-judge poolis also a list, but Arena runs one judge. It tries only entries from a family other than the parent's, in list order, so order it by preference. Aninherit-parententry never judges. Users can choose other detected models or panel sizes.
Show every role with its proposed value. Confirm with question, offering the detected models plus inherit-parent.
4. Validate
Every value must be a model from step 1, optionally with one of its listed variants, or inherit-parent. If one is not, ask again.
5. Write
Call pstack_roles with budget and the full roles map, panels as arrays. The budget value must be exactly unlimited, large, medium, or small, without a label or description. Pass reset: true to drop every stored value first.
6. Confirm
Tell the user the roles are stored and apply to the next subagent spawn.
7. Offer a verification skill (optional)
If the project has no verify-* skill or harness that drives the real app, offer once to create one with pstack-create-verification-skill. On no, move on.
8. Offer a formal verification skill (optional)
If the project has no formal verification skill (a project skill that runs a model checker, a proof, or another verifier against the code), offer once to create one with pstack-create-formal-verification-skill, so agents can prove a key property holds in every case in scope, not only the cases tests reach. On no, move on.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingoverlay/skills/setup-pstack/SKILL.md
Popularity
- Repository stars
- 1
- 7-day trend score
- 0
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/nexxeln/oc-pstack/setup-pstack)