pstack-opencode

A vendor-agnostic port of pstack by Lauren Tan (poteto) for OpenCode.

pstack is a set of engineering skills and principles that turn your coding agent into a disciplined engineering team. This port adapts pstack for OpenCode's tool system, removing Cursor-specific dependencies.

What's included

Component	Count	Description
poteto-mode	1	Main routing skill
principles/	21	Engineering principle skills
unslop	1	Auto-loading writing cleanup skill

Installation

Option 1: Copy to global skills

cp -r poteto-mode ~/.config/opencode/skills/
cp -r principles ~/.config/opencode/skills/
cp -r unslop ~/.config/opencode/skills/

Option 2: Copy to project skills

mkdir -p .opencode/skills
cp -r poteto-mode .opencode/skills/
cp -r principles .opencode/skills/
cp -r unslop .opencode/skills/

Option 3: Clone and symlink

git clone https://github.com/dsam/pstack-opencode.git ~/.config/opencode/skills/pstack-opencode
ln -s ~/.config/opencode/skills/pstack-opencode/poteto-mode ~/.config/opencode/skills/poteto-mode
ln -s ~/.config/opencode/skills/pstack-opencode/principles ~/.config/opencode/skills/principles
ln -s ~/.config/opencode/skills/pstack-opencode/unslop ~/.config/opencode/skills/unslop

Usage

Auto-loading unslop

The unslop skill has alwaysApply: true, so it should auto-load at session start. If it doesn't, explicitly load it:

skill(name="unslop")

Using poteto-mode

The main routing skill:

skill(name="poteto-mode")

Using individual principles

Each principle is a standalone skill:

skill(name="principle-laziness-protocol")
skill(name="principle-model-the-domain")
skill(name="principle-prove-it-works")

Principles

Core

Principle	When to apply
Laziness Protocol	Refactoring, sizing diffs, tempted to add abstractions
Foundational Thinking	Before writing logic
Redesign from First Principles	Integrating new requirements
Subtract Before You Add	Sequencing additions or refactors
Minimize Reader Load	Reviewing hard-to-trace code
Outcome-Oriented Execution	Planned rewrites and migrations
Experience First	Product/UX tradeoffs
Exhaust the Design Space	Novel decisions with no precedent
Build the Lever	Non-trivial work

Architecture

Principle	When to apply
Model the Domain	Stateful logic, branching code
Boundary Discipline	Validation, error handling, adapters
Type System Discipline	Designing types or signatures
Make Operations Idempotent	Commands, lifecycle steps, retries
Migrate Callers Then Delete	Introducing new internal APIs
Separate Before Serializing	Concurrent actors on shared state

Verification

Principle	When to apply
Prove It Works	After completing a task
Fix Root Causes	Debugging
Sequence Work into Verifiable Units	Multi-step work, stacking commits

Delegation

Principle	When to apply
Guard the Context Window	Large outputs, long files
Never Block on the Human	Reversible decisions

Meta

Principle	When to apply
Encode Lessons in Structure	Writing same instruction repeatedly

OpenCode mappings

pstack (Cursor)	OpenCode
AskQuestion	question tool
apply_patch	write / edit tools
poteto-agent	task(subagent_type="general")
Cursor models	OpenCode categories
cursor-team-kit	OpenCode built-in skills

Philosophy

pstack's core philosophy translates to any coding agent:

1. Write less, but higher quality code
2. Fearless parallelism — trust agents to write good, verifiable code
3. Rigorous engineering principles — not suggestions, but mandatory workflows
4. Verified work — prove it works before declaring done

Credits

• Lauren Tan (poteto) — Original pstack author
• Cursor Plugins — Original pstack repository
• OpenCode — The coding agent this port targets