Ailloy: The Package Manager for AI Instructions — A Helm-style Tool Built for Agent Workflows

Ailloy is an AI instruction package manager built for agent workflows, analogous to how Helm manages Kubernetes applications. It addresses problems such as messy version control of AI instructions, repeated configurations, difficulty in cross-project reuse, and poor tool compatibility. It introduces core concepts of Mold (configurable AI instruction packages) and Flux (configuration variables), supports multiple AI programming tools like Claude Code and Cursor, and provides a Helm-style command system to enable the creation, compilation, packaging, distribution, and reuse of AI instructions.

With the popularity of AI programming assistants and agent workflows, developers have accumulated a large number of AI instruction files, but face challenges like messy version control, repeated configurations, difficulty in cross-project reuse, and poor tool compatibility. Drawing on Helm's successful experience in managing Kubernetes applications, Ailloy aims to build a complete package management ecosystem for AI instructions.

Core Concepts:

• Mold: Analogous to Helm Chart, it is a versioned, configurable AI instruction package containing instruction templates, configuration files, and metadata;
• Flux: Analogous to Helm Values, it is a set of variables for configuring Mold, enabling "create once, use anywhere".

Helm-style Commands: Uses command names familiar to developers to reduce learning costs. For example, ailloy forge corresponds to helm template (dry-run rendering), ailloy cast corresponds to helm install (installing Mold), and new commands like ailloy anneal (interactive Flux configuration) and ailloy assay (lint check) are added.

Tool Agnosticism: Supports AI tools based on file instructions such as Claude Code, Cursor, Windsurf, and GitHub Copilot, and achieves multi-tool adaptation through flux.yaml mapping configuration.

Workflow Pipeline:

1. Creation: Write instruction templates using Go text/template;
2. Configuration: Use ailloy anneal to set Flux variables interactively;
3. Preview: Use ailloy forge for dry-run rendering verification;
4. Installation: Use ailloy cast to compile and install to the project (supports options like global installation);
5. Packaging: Use ailloy smelt to package as a tarball or binary;
6. Verification: Use ailloy temper to check the structural integrity of Mold.

Remote Mold: Supports direct installation of Mold from Git repositories (e.g., ailloy cast github.com/nimble-giant/nimble-mold@v0.1.10), and ailloy foundry manages the Mold registry (search, add, update, etc.).

Blank Types: Divided into three categories—commands (explicit calls like /brainstorm), skills (context-triggered), and workflows (GitHub Actions).

Ingot: Reusable template components that support cross-Mold sharing; bidirectional command syntax (e.g., ailloy foundry search is equivalent to ailloy search foundry); Claude Code plugin generation (e.g., ailloy plugin generate).

Ailloy is suitable for:

• Team Collaboration: Maintain standard AI instruction Mold to ensure consistent workflows among team members;
• Open-source Sharing: The community creates and shares common Mold (e.g., code review templates);
• Project Initialization: Quickly install pre-configured Mold to set up an AI development environment;
• Version Management: Controlled updates of AI workflows via versioned Mold to reduce regression risks.

Ailloy introduces software package management concepts into the AI workflow domain, solving key issues in AI instruction management. As AI programming assistants become popular in enterprises, Ailloy is expected to become an AI engineering infrastructure, helping teams better manage and expand AI-assisted development practices.