superhuman

Decompose requests into dependency-graph tasks with parallel wave execution and test-driven verification.

Updated Feb 19, 2026
One-click install
npx skills add https://github.com/brunolimaff-jpg/NOVO-APP --skill superhuman
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: superhuman
Source: https://github.com/brunolimaff-jpg/NOVO-APP/tree/main/.agents/skills/archive/2026-04-curation/superhuman
Command: npx skills add https://github.com/brunolimaff-jpg/NOVO-APP --skill superhuman

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Superhuman replaces traditional, meeting-heavy SDLC with an AI-native workflow that decomposes your request into a dependency-graph of tasks, runs them in parallel waves, and verifies each step with tests before proceeding.

Core Features & Use Cases

  • Dependency-graph decomposition: Breaks a feature into atomic sub-tasks with explicit dependencies (a DAG), avoiding rework and ordering bugs.
  • Wave-based parallel execution with TDD: Executes independent tasks in parallel waves while writing tests first (red/green/refactor) and only marking tasks done after verification.
  • Persistent progress board with resume: Tracks status in a project-root board file under .superhuman/board.md so you can resume midstream safely.
  • When it fits: End-to-end feature builds, multi-file refactors, greenfield work, and migrations.
  • When it does not: Quick single-file edits, typo fixes, or simple Q&A where deep planning and execution would be overkill.

Quick Start

Tell your AI: "Using superhuman, plan and build this feature end-to-end for my project, and include tests and verification before execution."

Frequently Asked Questions about superhuman

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I plan and execute multi-step feature builds using a dependency graph?

Multi-step feature builds use a dependency graph to decompose requests into atomic sub-tasks, assigning them to parallel execution waves while enforcing test-driven verification at every step to prevent ordering bugs and rework.

What is the best way to manage an AI-native software development lifecycle for complex refactors?

An AI-native software development lifecycle manages complex refactors by constructing a dependency graph of tasks, requiring an approval-first planning gate before file changes, and tracking progress in a persistent board file for safe resumption.

How does test-driven verification work during parallel task execution?

Test-driven verification during parallel task execution writes tests first using red/green/refactor cycles, only marking independent tasks as done after they pass verification within their assigned execution wave.

When should I avoid using dependency-graph task decomposition for software development?

Dependency-graph task decomposition should be avoided for quick single-file edits, typo fixes, or simple Q&A, where deep planning, parallel execution waves, and test-driven validation introduce unnecessary overhead.

How do I resume a multi-file migration project that was interrupted midstream?

To resume an interrupted migration midstream, the system reads a persistent progress board file stored in the project root, allowing safe continuation of the remaining dependency-graph tasks from their last verified state.

Does wave-based parallel execution require an approval step before modifying files?

Wave-based parallel execution requires an explicit approval-first planning gate after dependency graph construction and before any file changes are made, ensuring all task ordering and decomposition logic is validated.