dmi-superpowers

dmi_superpowers

An open-source, continuously distilled collection of the best skills and concepts for giving coding agents the discipline to write high-quality code — consolidated from the superpowers plugin and Matt Pocock's skills into a single coherent methodology engine.

Contributions are welcome. See AGENTS.md for what belongs here and how to write a pull request.

What It Is

dmi_superpowers packages 25 skills that guide coding agents through a disciplined workflow:

brainstorm → PRD → grill-with-docs → TSP → grill → TDD → review → finish

The agent never just starts coding. It steps back, refines requirements, validates against domain language, plans carefully, writes tests first, reviews its own work, and closes out cleanly.

The 25 Skills

Kept from superpowers (core workflow)

Skill	Purpose
brainstorming	Socratic design refinement before any code
writing-plans	Detailed implementation plans with task breakdown
executing-plans	Batch execution with human checkpoints
dispatching-parallel-agents	Concurrent subagent workflows
requesting-code-review	Pre-review checklist and dispatch
receiving-code-review	Responding to feedback with rigor
resolving-merge-conflicts	Resolve an in-progress git merge/rebase conflict
using-git-worktrees	Parallel development branches
finishing-a-development-branch	Merge/PR decision workflow
creating-a-pull-request	PR title/body for non-developer + developer audiences
subagent-driven-development	Fast iteration with two-stage review
using-superpowers	Introduction to the skills system
writing-skills	Create new skills following best practices

TDD-swapped (superpowers TDD replaced with tdd skill)

Skill	Purpose
test-driven-development	RED-GREEN-REFACTOR cycle (from tdd skill)
systematic-debugging	4-phase root cause process
verification-before-completion	Ensure it's actually fixed

Added from Matt Pocock's skills

Skill	Purpose
codebase-design	Shared vocabulary for designing deep modules — interfaces, seams, depth
domain-modeling	Build and sharpen the domain model; record architectural decisions (ADRs)
grill-with-docs	Challenge plans against CONTEXT.md and ADRs
grilling	Relentlessly interview to stress-test a plan or design before building
improve-codebase-architecture	Find deepening/refactor opportunities
say	Communicate code findings in plain + technical pairs, and rewrite prose for human readers
to-prd	Convert rough ideas to structured PRDs

Merged / unified (debugging consolidated)

Skill	Purpose
prototype	Rapid prototyping skill

Utility / mode skills

Skill	Purpose
ponytail	persistent minimalism mode: defaults every solution to the simplest thing that works (YAGNI, stdlib/native first, shortest diff), at intensity lite/full/ultra. Off by default; type ponytail [level] or be lazy to activate (plain text, not a slash command — ADR-0005).

Installation

Install dmi-superpowers separately for each harness you use.

Claude Code

/plugin install dmi-superpowers@dmi-superpowers-marketplace

Or from this repository directly:

/plugin marketplace add HundredBillion/dmi_superpowers
/plugin install dmi-superpowers@dmi-superpowers-marketplace

Antigravity

agy plugin install https://github.com/HundredBillion/dmi_superpowers

Codex CLI

• Open the plugin search interface:

  /plugins
  

• Search for dmi-superpowers and select Install Plugin.

Gemini CLI

gemini extensions install https://github.com/HundredBillion/dmi_superpowers

Update later:

gemini extensions update dmi-superpowers

Kimi Code

/plugins install https://github.com/HundredBillion/dmi_superpowers

Detailed docs: docs/README.kimi.md

OpenCode

Fetch and follow instructions from https://raw.githubusercontent.com/HundredBillion/dmi_superpowers/refs/heads/main/.opencode/INSTALL.md

Detailed docs: docs/README.opencode.md

Pi

pi install git:github.com/HundredBillion/dmi_superpowers

For local development:

pi -e /path/to/dmi_superpowers

The Workflow

1. brainstorming — Activates before writing code. Refines rough ideas through questions, explores alternatives, presents design in sections for validation. Saves design document.

2. to-prd — Converts conversation output into a structured PRD. Produces docs/PRDs/.

3. grill-with-docs — Challenges the PRD against CONTEXT.md and ADRs. Sharpens terminology, updates domain docs.

4. writing-plans — Breaks work into bite-sized tasks (2–5 min each). Every task has exact file paths, interfaces, and verification steps. Produces docs/TSPs/.