explaingit

wind7rui/hyperspec

24Audience · developerComplexity · 3/5ActiveSetup · easy

TLDR

An orchestration skill that wires OpenSpec and Superpowers into a three-stage propose-apply-archive workflow for Claude Code, Cursor, and Codex CLI, with state tracked in YAML.

Mindmap

mindmap
  root((HyperSpec))
    Inputs
      User request
      Project files
      Spec library
    Outputs
      Proposal docs
      Implementation plan
      Git commits
      State YAML
    Use Cases
      Spec driven coding
      Stage gated workflow
      Auto build detection
      TDD with subagents
    Tech Stack
      Claude Code
      Cursor
      Codex CLI
      OpenSpec
      Superpowers

Things people build with this

USE CASE 1

Run a feature from vague request through spec, implementation, and archive in one tool

USE CASE 2

Force a coding assistant to stay in the right stage and stop writing code before specs exist

USE CASE 3

Auto-detect Maven, npm, Go, Cargo, or pyproject and pick build commands

USE CASE 4

Track multi-session work through .hyperspec-state.yaml without re-prompting

Tech stack

ClaudeCodeCursorCodexCLIOpenSpecSuperpowersYAML

Getting it running

Difficulty · easy Time to first run · 30min

You must also install OpenSpec and Superpowers since HyperSpec is a pure orchestration layer that calls those skills.

In plain English

HyperSpec is an add-on for AI coding assistants like Claude Code, Cursor, and Codex CLI that ties together two other add-ons into one disciplined workflow. The first is OpenSpec, which manages specification documents: what you are going to build and why. The second is Superpowers, which manages test-driven development, planning, and code review by subagents. HyperSpec sits on top of both and does only four things: it senses what kind of project you have, checks current progress, routes the work to the right stage, and enforces a commit habit. The author is careful about scope. HyperSpec is described as a pure orchestration layer. It does not rewrite anything that OpenSpec or Superpowers already do. It does not create spec artifacts by hand, it does not convert tasks into plans by hand, and it does not run the archive step by hand. It calls those existing skills at the right moment instead. The same idea applies in reverse: it refuses to write code in the spec stage and refuses to edit specs in the implementation stage. A full run is split into three stages. The propose stage turns a vague user request into a proposal, a design document, a spec change set, a task list, and a step by step implementation plan with checkboxes. The apply stage executes that plan, picking between a full mode with subagents and a lightweight mode based on factors like task count and how many modules are touched. After each task it compiles, ticks the checkbox, updates state, and commits, but never pushes. The archive stage checks that the code matches the spec, calls the archive skill to merge the spec into the main spec library, and tidies the branch. Progress is tracked in a file called .hyperspec-state.yaml, and HyperSpec always verifies that file against the real project layout so a stale cache cannot mislead it. The project sensor reads file extensions and config files like pom.xml, package.json, go.mod, Cargo.toml, and pyproject.toml to pick the right build and test commands automatically. Installation is a copy of the folder into the skills directory of the editor you use.

Copy-paste prompts

Prompt 1
Walk me through the propose, apply, and archive stages of HyperSpec and which skills get called in each
Prompt 2
Show how the project sensor in HyperSpec maps pom.xml or go.mod to a build and test command
Prompt 3
Explain why HyperSpec refuses to edit specs during the apply stage and how it enforces that
Prompt 4
Modify HyperSpec so the apply stage also runs an external linter before each commit
Prompt 5
Write a sample .hyperspec-state.yaml that shows a half-finished apply stage on a Cargo project
Open on GitHub → Explain another repo

Generated 2026-05-22 · Model: sonnet-4-6 · Verify against the repo before relying on details.