Skip to content

Repository files navigation

Hope Protected Light icon

Hope

Hope helps people work with AI while staying able to see, understand, and control the work.

한국어


Features

🤝 Align — Reach shared understanding before implementation and prevent intent debt

Align reviews the request against verifiable evidence and interviews the person about material gaps, contradictions, risks, unsupported assumptions, edge cases, and simpler alternatives that could change the result.

When the agreement needs a durable record for later work or review, or the person asks for an artifact, Align writes one self-contained HTML brief inside the project. A small, clear task that will continue in the current session can stay in the conversation without creating a file.

A brief contains one agreed goal and a set of completion criteria. Each criterion includes how it is verified and who—AI or the person—judges it. Material changes remain as new versions in the same file, and the document serves as the implementation contract.

For material UI work without attached references, Align checks the project first, uses web search when needed, and presents two or three image mockups.

Important

Generated Align briefs are project documentation. Later version-control work includes them with related project changes unless the person excludes them.

Complete example HTML: Open the English Align brief for a fan schedule that makes source conflicts, changes, cancellations, and verification ownership explicit.

The captures below come from this example. It uses sample data and does not represent the live rescene.fan interface.

English Hope Align example showing the trusted fan-schedule goal, completion criteria, and boundary

View detailed Align captures
Compared UI directions Source and lifecycle decisions
Two UI directions for the trusted fan schedule in an English Align artifact Source priority, conflict, freshness, and cancellation decisions in an English Align artifact

🔎 Diff — Understand what changed and how to judge it to prevent cognitive debt

A code change can be complete while its owner still cannot predict, explain, or judge it, and that gap is cognitive debt.

Diff creates one HTML artifact that explains behavior before code and links important claims to evidence.

It may use visuals, a microworld, or a quiz to help the reader explore the change.

The artifact helps the reader understand and judge the change, then use that understanding in follow-up decisions and work.

Note

With no URL, Diff first looks for the current branch's pull request. If none exists, it selects your latest open pull request in the repository. Run Diff again when the pull request changes.

The captures below come from a fixed English Diff example based on Ky PR #867.

Complete example HTML: Open the English Diff artifact for Ky PR #867 with its retry-configuration microworld and quiz.

English Hope Diff example for Ky pull request 867 showing the goal, before and after behavior, and verification item

View detailed Diff captures
Core change Interactive microworld
The retry configuration change explained in an English Diff artifact An interactive retry-configuration microworld in an English Diff artifact

An understanding quiz about retry behavior in an English Diff artifact


⚖️ Toxic Review — Put a work product through a rigorous Red–Blue review

Red finds. Blue challenges. The active agent judges.

Independent Red reviewers probe distinct material risks. When a finding is consequential or materially uncertain, a fresh Blue verifier sees only the sealed finding and scoped evidence. Blue tries to disprove it and expose overstatement or missing context; it does not defend the work product or decide the result.

The active agent retains final judgment and reports only the findings supported by the evidence.

Tip

Ask Hope to limit the Red reviewer count when you want a smaller routine run. Review size alone does not add Blue, but a consequential or materially uncertain finding still triggers it.


✨ Polish — Refine implemented work

Independent review agents look for useful improvements.

For code, they check reuse of existing helpers, simplicity, efficiency, and abstraction fit.

A fresh finisher judges the results, applies only the improvements that work together, and verifies the result.

Polish does not hunt for bugs, develop features, perform migrations, or handle broad maintenance.


🧹 Sweep — Clean up a codebase

Sweep performs a read-only review of a codebase.

It looks for broken references, stale code, unsupported abstractions, verification gaps, dependency or license risk, delivery waste, unclear ownership, and similar maintenance risks.

Select a candidate from the review results to start work.


✍️ Write — Make language clearer without losing meaning

Hope also uses Write within other tasks, including implementation and other Skills.

Write's shared standard adapts George Orwell's six rules in Politics and the English Language.


Install

You need:

  • Node.js 22 or newer
  • An authenticated GitHub CLI to use Diff. Run gh auth login first if needed.

Tip

The simplest option is to ask an AI:

Install Hope from https://github.com/dkstm95/hope for this host.
Follow the repository README and tell me if I need to restart.

To install it yourself, run the commands for your host.

For example:

# Codex
codex plugin marketplace add dkstm95/hope
codex plugin add hope@hope
# Claude Code
claude plugin marketplace add dkstm95/hope
claude plugin install hope@hope

License

MIT

Releases

Packages

Contributors

Languages