FF

freenet/freenet-agent-skills

开发工具

本页由 AI 翻译,技术细节请以原文为准。 查看原文

27 stars 质量 67 趋势 31LGPL-3.0

在 Freenet 上构建应用程序的 AI 编码代理技能

概览

在Freenet上构建应用程序的AI编码代理技能。兼容Claude代码和OpenCode。 在 Freenet 上构建去中心化应用。指南内容:1. 设计契约(共享状态,复制状态) 2.实现代理(私有,本地状态) 3.构建界面(WebSocket 连接 Freenet) 基于River,一个去中心化聊天应用。 创建高质量Freenet拉取请求的指南。包括:- 四个并行审查子代理(代码优先、测试、怀疑、大局)- 大局审查捕捉“CI追逐”反模式 - 测试质量标准与回归预防 - 基于工作树的工作流 调试非平凡问题的方法论:- 代码变更前的假设形成 - 与子代理的并行调查 - 应避免的反模式(跳到结论,削弱测试) - 测试覆盖率差距分析 风险分级、多视角的公关审查——对变更进行分诊,并行运行专业子代理(代码优先,...

概述(AI 翻译)

在Freenet上构建应用程序的AI编码代理技能。兼容Claude代码和OpenCode。

可用技能

Dapp-builder

在Freenet上构建去中心化应用。以下指南:

  1. 设计合同(共享状态,复制状态)
  2. 执行代表(私人、地方州)
  3. 构建用户界面(WebSocket连接到Freenet)

基于River,一个去中心化聊天应用。

创作

创建高质量Freenet拉取请求的指南。包括:

  • 四个并行审查子代理(代码优先、测试、怀疑、宏观)
  • 《宏观回顾》发现“CI追逐”反模式
  • 测试质量标准与回归预防
  • 基于工作树的工作流

系统调试

调试非平凡问题的方法论:

  • 在代码变更前形成假设
  • 与亚代理的平行调查
  • 应避免的反模式(跳到结论,削弱测试)
  • 测试覆盖率差分析

公关评论

风险分级、多视角的公关审查——对变革进行分诊,并派遣专门的分级代理人 并行(代码优先、测试、怀疑、宏观),加上一个外部模型通过,缩放到

本页由 AI 翻译,技术细节请以原文为准。

README

以下为原文 README 节选,完整内容请查看原文。

Freenet Agent Skills

AI coding agent skills for building applications on Freenet. Compatible with Claude Code and OpenCode.

Available Skills

dapp-builder

Build decentralized applications on Freenet. Guides through:

  1. Designing contracts (shared, replicated state)
  2. Implementing delegates (private, local state)
  3. Building the UI (WebSocket connection to Freenet)

Based on River, a decentralized chat application.

pr-creation

Guidelines for creating high-quality Freenet pull requests. Includes:

  • Four parallel review subagents (code-first, testing, skeptical, big-picture)
  • Big-picture review catches “CI chasing” anti-patterns
  • Test quality standards and regression prevention
  • Worktree-based workflow

systematic-debugging

Methodology for debugging non-trivial problems:

  • Hypothesis formation before code changes
  • Parallel investigation with subagents
  • Anti-patterns to avoid (jumping to conclusions, weakening tests)
  • Test coverage gap analysis

pr-review

Risk-tiered, multi-perspective PR review — triages the change, runs specialist subagents in parallel (code-first, testing, skeptical, big-picture), plus an external model pass scaled to risk, then posts a consolidated review to the PR.

release

Orchestrates a Freenet release. It runs the mandatory pre-cut gates first: no concurrent cut, the auto-update updater diff and canary suites, a 30-minute real-NAT soak of the exact ship SHA, and a telemetry baseline. Then it determines the next version, shows the changelog, confirms with the user, triggers the release pipeline via GitHub Actions, and monitors telemetry through the 0-4h auto-update stagger.

linux-test

Runs integration tests that require Linux’s full loopback range via Docker — for tests that fail on macOS with “Can’t assign requested address”.

local-dev

Set up and manage local Freenet development environments: run a local node, publish contracts, query connections/diagnostics, and iterate on a Freenet application without deploying to the live network.

Hooks

The hooks/ directory contains two types of hooks for Freenet development:

1. Claude Code Hooks (Automatic)

File: hooks/hooks.json

Automatically runs cargo fmt and cargo clippy before Claude runs git commits. Included with the plugin - no setup required.

2. Git Pre-Commit Hook (Manual)

File: hooks/pre-commit

Runs cargo fmt and cargo clippy before git commits. Must be installed manually:

# From freenet-core repository
curl -fsSL https://raw.githubusercontent.com/freenet/freenet-agent-skills/main/hooks/pre-commit > .git/hooks/pre-commit
chmod +x .git/hooks/pre-commit
在 GitHub 查看完整 README

推荐工具

换一个关键词,或者移除筛选条件。

安装

npx skillfish add freenet/freenet-agent-skills
新手?先看 mcpskill.org 教程 →