dsh-claude-delegate：在 DSH 中把自包含编码子任务委派给本机 Claude Code
dsh-claude-delegate 是 project-hy 维护的 DSH 工作流插件，通过官方 Claude Agent SDK 驱动本机 Claude Code CLI，注册工具 claude_code 与监控面板。支持后台作业、会话 resume、长命令实时输出通道 mcp__dsh__shell、JSON Schema 结构化结果与成本统计。安装命令为 dsh plugin --profile <profile> add dsh-claude-delegate，需本机已登录 claude CLI 并重启 DSH。适合在 DSH 中编排、把边界清晰的编码子任务委派出去的场景；使用前请审阅源码与权限影响。
Read Moreae-workflow：为 Claude Code 与 Open Code 准备的双端 Agent 工作流 Skill
ae-workflow 是 SkillHub 社区 Skill（@user_df62fd7e/ae-workflow，v4.0.0），为 Claude Code 与 Open Code 分别提供工作流，分类 ai-agent，标签 requires_api_key=false。文章说明其定位、可核实功能，以及按 skillhub install 与各客户端 skills 目录（如 ~/.claude/skills/、~/.cursor/skills/）的安装方式；典型用法为安装后由 Agent 读取 SKILL.md 分端执行，并提示社区来源与页面未公开逐步细节等注意点。
Read MoreBoo哥AI流程图-drawio：用 Skill 生成可编辑的 .drawio 图表
Boo哥AI流程图-drawio（@user_3c9003af/boo-ai-drawio，v1.0.0）是 SkillHub 上的 design-media 类社区 Skill，由 Boo哥AI智写发布。它面向 Claude Code 等支持 Skill 的客户端，通过交互模式生成原生 .drawio 文件，并可导出 PNG、SVG、PDF、JPG、WebP；提供 7 类图表、6 套主题与三档复杂度，并支持中文排版。本文基于 SkillHub 页面说明安装路径（skillhub install 与 skillhub.cn/install/skillhub.md）、典型自然语言用法及使用注意，无需 API Key。
Read More虚幻引擎 Claude Code 技能：为 Agent 补齐虚幻引擎协作上下文
本文介绍 SkillHub 社区技能「虚幻引擎 Claude Code 技能」（@user_3c6cb52e/unreal-engine-sh-etzk7s，v1.0.0，ai-agent，requires_api_key=false）。它面向在 Claude Code、Cursor 等客户端中用 Agent 处理虚幻引擎相关任务的场景，将协作约定封装为通用 SKILL.md 格式。文中给出 skillhub CLI 检查、按客户端 skills 目录安装命令、以及官方页面提供的 Agent 安装提示句，并说明能力边界以安装后的 SKILL.md 为准，避免臆造未公开的功能细节。
Read MoreSkill 设计师：把已验证路径复刻为可安装的 Skill
Skill 设计师（@user_00c9b356/chenlin-skillcreator，v3.3.0）是 SkillHub 社区 dev-programming 类 Skill，用于从零创建或迭代 Skill。融合六大设计原则、Loop Thinking 与三层分离，支持任务执行、习惯养成、工具使用三类，产出 SKILL.md、REQUIREMENTS.md、SOLUTIONS.md、MISTAKES.md。本文介绍定位、能力、SkillHub CLI 安装路径与典型对话用法，并说明何时不应误用于仅调用已有 Skill 的场景。
Read More使用 Nature Skills 运行自然风格学术写作与图表工作流程
SkillHub 社区 Skill「使用 Nature Skills 运行自然风格的学术写作和图表工作流程」（v1.0.0）将手稿润色、学术写作、引用、论文阅读、回复起草、数据处理、幻灯片与科学图表等能力打包为可重用的 Codex 或 Claude Code 技能包，且标注 requires_api_key=false。本文依据 SkillHub 页面与安装文档说明坐标、功能边界，以及通过 Agent 提示或 skillhub install --dir 安装到 ~/.cursor/skills/ 等目录的步骤与注意事项。
Read More归藏PPT制作技能：为 AI 编程助手接入 PPT 领域能力
归藏PPT制作技能是 SkillHub 社区版 office-efficiency 技能（v1.0.0，坐标 @user_3c6cb52e/guizang-ppt-skill-png），为 AI 编程助手提供 PPT 相关领域能力，支持 Claude Code、Codex、WorkBuddy、OpenClaw、QClaw、Hermes Agent，且 requires_api_key=false。文章说明其定位、可核实功能边界，并给出按 skillhub.cn/install/skillhub.md 安装及 skillhub install --dir 指向各 Agent skills 目录的步骤；典型用法以安装后阅读 SKILL.md 为准。
Read Morerust-review: A Security Review Skill for Rust Codebases
rust-review is a Rust code security review Skill in the Trail of Bits skills marketplace. It uses an orchestrator with worker, dedup, and fp subagents to cover safe/unsafe boundaries, unsafe memory safety, concurrency, panic-DoS, FFI, and async runtime issues, producing REPORT.md and REPORT.sarif. It is suitable for auditing crates and services that include unsafe code, FFI, or concurrency; for scenarios such as pure C/C++, smart contracts, and key zeroing, use c-review, contract-specific Skills, or zeroize-audit instead. After adding the trailofbits/skills marketplace in Claude Code or Codex, start the review with /rust-review:rust-review.
Read Moreskill-improver: Polish SKILL.md to Publishable Quality with a Fix-Review Loop
skill-improver is a meta-skill in the Trail of Bits skills marketplace. It depends on skill-reviewer from plugin-dev and runs iterative review-and-repair cycles on a target SKILL.md, remediating issues by Critical, Major, or Minor severity and re-reviewing them until it emits skill-improvement-complete or reaches the default limit of 20 iterations. It is well suited to batch-aligning frontmatter, trigger descriptions, and When to Use structure; for a single review, use skill-reviewer directly. To install it, first add the trailofbits/skills marketplace and enable plugin-dev, then start the loop in Claude Code by running /skill-improver with the target path.
Read Morecontext-fundamentals: Foundational Mental Models for Context Engineering
context-fundamentals is the foundational Skill in muratcankoylan's open-source collection, Agent Skills for Context Engineering. It uses SKILL.md to define the full composition of context during inference (system prompt, tool definitions, retrieved documents, message history, and tool outputs), and establishes mental models such as a finite attention budget and the U-shaped attention curve, emphasizing signal quality over token accumulation. The article explains when this Skill should be activated, its routing relationship with operational Skills such as context-degradation, and directory-level installation methods via the Claude Code plugin marketplace, Cursor, and Codex. It also provides an example of the official system prompt partitioning and a recommended reading order, making it suitable as introductory background reading for the context engineering series.
Read Moreagent-patterns: Multi-Agent Design Patterns and Orchestration Practices
agent-patterns is a Multi-Agent-focused skill in the oakoss agent-skills collection (112 hand-curated skills; v1.1, MIT). It catalogs patterns such as fungible swarm, pipeline, hierarchical orchestration, and two-stage review, and explains Claude Code subagent configuration paths, tool allowlists, guidance on parallelism and batch processing, and constraints such as subagents not being nestable. It is suitable as a reference when designing custom .claude/agents, delegating batch tasks, or debugging coordination issues; it can be installed with pnpm dlx skills add oakoss/agent-skills -s agent-patterns.
Read MoreDeepSeek Harness Plugin: dsh-plugin-ponytail
`dsh-plugin-ponytail` is a plugin that integrates Ponytail's "laid-back senior developer mode" into DeepSeek Harness (DSH). It injects Always-on rules through `systemPrompt.context`, provides skills and a `/ponytail` command interface to control the mode intensity, and uses the `.ponytail/` directory and `state` file to enable project-level activation and state persistence. Unlike similar plugins that take effect globally or provide only skills, this plugin takes effect only when the project root contains that directory, and is installed as a mutually exclusive option.
Read MoreDeepSeek Harness Plugin: Codex and Claude Code Subagent Presets
The `dsh-preset-codex-claude` plugin enables DeepSeek Harness (DSH) to integrate Codex and Claude Code through official subagents. It supports read/write separation and parallel execution, and provides two permission modes: full access and read-only. Installation first requires installing dependency packages into the DSH root directory, then importing them via a Profile. The plugin provides four tools (such as `subagent_codex`), and subagents run as independent processes without inheriting session context. When using the plugin, attention should be paid to version compatibility, process-level caching, and session isolation issues.
Read Moredsh-auto-memory: A Claude Code–Style Automatic Memory Plugin for DeepSeek Harness
`dsh-auto-memory` is a plugin that provides persistent memory capabilities for DeepSeek Harness (DSH). It uses pure file-based storage to support typed memories, scope isolation, and data security mechanisms, addressing the issue of agents resetting memory in new sessions. The plugin automatically injects memory indexes and features atomic writes, truncation prevention, and zero external dependencies, making it suitable for developers who need to maintain long-term project context and coherence.
Read Moredsh-init: Claude Code-style /init Command for DeepSeek Harness
The `dsh-init` plugin, officially maintained by DeepSeek Harness (DSH), brings Claude Code’s `/init` workflow into DSH to address the pain point of cumbersome background context setup when building agents. Through a seven-stage workflow, the plugin automatically generates project-level guidance in `CLAUDE.md` and uses symbolic linking to create `AGENTS.md`, enabling multiple tools to share the same single source of truth. After installation, restarting DSH allows the workspace to be initialized via slash commands or natural language, helping quickly establish project context.
Read MoreDeepSeek Harness Claude Code Integration Plugin
`relay-dsh-plugin-claude` is an open-source plugin designed to integrate Claude Code as a native backend into the DeepSeek Harness (DSH) Web UI. It addresses context fragmentation caused by developers switching between DSH and Claude Code, supports seamless integration, and preserves DSH's native experience, session continuity, and hybrid tool support. Installation requires a prepared Node.js environment and completed local authentication. The plugin does not collect credentials and does not include Relay Events, file browsing, or terminal panel functionality.
Read Moredsh-user-prompt: DeepSeek Harness Custom Prompt Plugin
The DeepSeek Harness (DSH) plugin dsh-user-prompt provides developers with a convenient prompt management feature. It supports editing and injecting Markdown-formatted global prompts directly from the settings page, eliminating the need to manually paste repeated background information. It is compatible with Claude Code and Codex global rules, supports one-click import and export, and, as a purely additive plugin, does not affect official features. Once configured, the content is injected into conversations through the context channel, but you should avoid using DSH template variable syntax.
Read Moredsh-subagent-claude-live: Claude Code Streaming Subagent
The DeepSeek Harness (DSH) plugin `dsh-subagent-claude-live` addresses limitations such as the invisibility of the built-in subagent runtime and inflexible model configuration. It uses the `claude_code` and `claude_models` tools to turn Claude Code into a streamable subagent with real-time trajectory viewing. The plugin supports specifying the model, effort level, and permission mode at runtime, and can retrieve a real-time model list. Installation requires DSH version >= 0.1.1-rc.1 and a system-installed `claude` CLI (or a bundled payload). Configuration allows defining aliases to pin the model and effort level. The main limitations are that each run spawns a new process, history resumption is not supported, and deep mandatory execution is not provided.
Read Moredsh-plugin-tui: A Claude Code-Style Terminal UI for DeepSeek Harness
`dsh-plugin-tui` is a client-side plugin for DeepSeek Harness, designed to provide a persistent, interactive terminal environment similar to Claude Code. The plugin supports streaming output, session management, model switching, and permission approval, addressing the limitation of traditional single-turn plugins in enabling multi-step collaboration. Users must install it by specifying the `tui` configuration file; the plugin requires only two bundle layers and does not support `dsh-headless`. Its core features include internal REPL commands (such as `/sessions` and `/model`) and real-time status feedback, though known considerations include one-time approval limitations and the need for token rendering optimization.
Read Moredsh-plugin-dew: View the Status of Other Coding Agents
The DSH plugin `dsh-plugin-dew` enables DSH Desktop to read local logs from agents such as Claude Code, Codex, and Cursor, and normalize their statuses into `needs you`, `running`, `done`, and `idle`. Installation requires Node.js >= 20 (23.8+ for reading compressed logs). Core features include querying session status via `agent_sessions` and scheduled tasks via `agent_scheduled_tasks`. The plugin runs in read-only mode only, without writing data or making network requests.
Read MoreWhale Girl Desktop Pet (whalegirl-deskpet)
Whale Girl Desktop Pet is an open-source desktop companion project supporting the DSH, Claude Code, and WorkBuddy platforms. The character assets are AI-generated, and the idle animations are programmatically baked to avoid jitter. Core features include: serving a bowl while work is in progress and clearing the bowl when a turn ends; pressing Esc triggers an interruption reaction and a “trashy” expression; clicking enables interaction; and while an agent is running, a small dolphin swims past its feet. Installation methods differ by platform, and the portable version is about 200MB. Note Windows security prompts and renderer compatibility across different versions.
Read MoreSpecWave: Multi-Host Coding Standards and Harness Workflow Tool
SpecWave is a multi-host coding CLI tool that natively integrates the ICVO specification into clients such as Cursor, Claude Code, and DSH through declarative adapter tables. It provides built-in Harness process commands and task-gating mechanisms to ensure consistent and verifiable coding behavior across tools. Users must meet the Node.js version requirements, apply configurations via the `host apply` command, and verify task gating using `verify`, establishing a strongly constrained execution environment for multi-agent collaboration.
Read Moredsh-claude-code: Running Claude Code in DeepSeek Harness
The `dsh-claude-code` plugin enables seamless switching to Claude Code within DeepSeek Harness (DSH) workflows while maintaining DSH's interface and logging capabilities. It defines a `claude-code` engine and works with `dsh-engine-switch` to provide one-click switching, real-time trace visibility, context usage monitoring, and permission bridging. Tools and sandboxing are directly provided by Claude Code, with precise continuation and the ability to retain DeepSeek subagents. Installation depends on `dsh-engine-switch` and a local Claude CLI, and the plugin can be used after configuration and restart.
Read Moredsh-plugin-local-agent-bridge: Control Claude Code and Codex in the Browser
`dsh-plugin-local-agent-bridge` is a client plugin that connects local Claude Code and Codex sessions to the DeepSeek Harness browser interface. It uses a local process bridge to keep authentication and execution on the host terminal while the browser serves only as the UI, avoiding frequent window switching. The plugin provides a unified panel, keyboard shortcuts, multiple permission modes, session monitoring, and native tool approval. Installation must meet specific version requirements and be launched through the built-in web profile to ensure credentials are not transferred.
Read Moredsh-skill-adapter: Importing Claude Code Skills into DeepSeek Harness
Maintained by zp-home, this plugin provides a toolchain for skill import, normalization, and distribution in the DeepSeek Harness (DSH) ecosystem. It bridges the migration gap from the Claude ecosystem and includes three core capabilities: 1. **Parsing and normalization**: zero-dependency parsing of SKILL.md and unified frontmatter conventions; 2. **CLI import tool**: supports batch fetching, validation, and compatibility report generation; 3. **Runtime mounting**: dynamically mounts GitHub or local skill sources via DSH plugins. Users can quickly install and enable it through the CLI or a configuration file, with Node.js 18.17+ support.
Read Moredsh-claude-code: Delegating Coding Subtasks to a Local Claude Code Subscription
The DeepSeek Harness plugin `dsh-claude-code` leverages the official Claude Agent SDK to integrate local Claude Code subscription capabilities into DSH, enabling task delegation and collaboration. It provides core features such as permission management, cross-turn memory, background asynchronous execution, and a web-based monitoring dashboard. Models, permissions, and agent settings can be adjusted through configuration options. It supports both foreground synchronous and background asynchronous execution and includes quota-viewing tools. Users should note dependencies on background tasks and the need for restarts to apply changes, making it suitable for local coding scenarios where cost control is important.
Read Moredsh-tui-en: A Claude Code–Style Interactive Terminal Interface for DeepSeek Harness
`dsh-tui-en` is a terminal-native UI plugin for DeepSeek Harness that delivers a Claude Code-style interactive experience. It addresses the lack of visualized context management and streaming feedback in terminal interactions, enhancing Agent state monitoring through streaming Markdown, tool cards, timeline navigation, and full session workflows. The plugin does not modify the DSH core, supports long-session design, is currently in Public Beta, and is suitable for daily development use.
Read MoreDeepSeek Harness Plugin: dsh-claude-code-templates
The DeepSeek Harness (DSH) plugin addresses the shortage of skill templates available to developers for configuration by porting 884 Claude Code skill templates. It contains 29 categories, covers official Anthropic skills, and the content is copied verbatim, byte for byte, so it can be invoked directly without additional configuration. After installing via a local path, users can load and use it directly in DSH. Note that the plugin repository is currently private, and some content has been removed due to licensing restrictions; verify the source code before installation to ensure compliance with usage policies.
Read Moredsh-btw — Add Native Claude Code /btw Side Questions + /remember Session Notes to DeepSeek Harness
The DeepSeek Harness plugin `dsh-btw` is designed to help developers quickly ask side questions and take notes within the main conversation. It includes two core slash commands: `/btw` for background side questions that do not interrupt the main flow and are not persisted to disk, and `/remember` for session memos with persistent storage support. The plugin is a zero-dependency bundle, installable via link, and supports configuration validation and data cleanup.
Read Moredsh-plugin-capabilities: Moving Skills and MCP Server Management into the dsh Settings Page
# Introduction When developing with dsh, skills and MCP server configurations are mostly stored in files: skills are individual `SKILL.md` files, and MCP servers are defined in YAML lines within profile patches or global patches. To temporarily disable a skill or modify a server parameter, you must leave the interface and manually edit files. If you are also using Claude Code or Codex, you would need to maintain a separate copy of the same set of skills and MCP configurations. `dsh-plugin-capabilities` solves these two problems: it allows you to manage skills and MCP servers directly within the dsh settings page and supports one-click import from other agents. The following sections introduce the features, installation, and typical usage in order.
Read Moredsh-inherit: One-click import MCP servers and Skills from Claude Code, Cursor, Codex, and cc-switch to DeepSeek Harn
dsh-inherit is a plugin (version 0.1.0, Developer preview, MIT license, unofficial) developed by community developer MayBeTheWorld for DeepSeek Harness (DSH). It enables one-click importing of MCP servers and Skills scattered across Claude Code, Cursor, Codex, and cc-switch into DSH.
Read Morejune-agent: June Dual-Personality Agent and Supporting Skills, installable as a DSH plugin
june-agent is an npm package maintained by lemonorangeapple that packages two fixed-personality Agents and two skills into a cross-hosting solution. June-M features a "catgirl" personality (temperature 0.1, obedient and concise), while June-S features a "dominant" personality (temperature 0.3, disciplinary pushback). The skills include Git backup strategies (git-backup) and design coding standards (june-style). The package provides both Claude Code format (agents/ and skills/ directories) and DeepSeek Harness format (presets/ directory), DSH.
Read Moredsh-claudecode-tips: Replace "Deep diving..." in DSH web profile with Claude Code style loading words
dsh-claudecode-tips (or @claude-code-tips/dsh-plugin) is a DSH plugin developed by ldgeng, designed for the web profile. It replaces the static status word "Deep diving..." during the turn runtime with the original 56 randomly switching loading words from Claude Code (such as "Thinking" and "Pondering"), while retaining the elapsed time timer. It is a pure browser-side client plugin that monitors the chat DOM.
Read Moredsh-simple-cli: Add a Claude Code-like terminal chat to DeepSeek Harness
# Foreword Developers who work in the command line for a long time often have this requirement: input a sentence directly in the terminal to see the model's response in a streaming format, the tool execution process in real-time, and manually confirm sensitive operations—instead of switching to a browser or assembling outputs from headless calls. DeepSeek Harness (hereinafter referred to as dsh) supports specifying a profile at startup (e.g., `dsh --profile cli`). The dsh-simple-cli introduced below is a CLI profile that runs entirely within the terminal: once installed, use the `dsh-cli` command to converse with the Agent, resembling the interface of Claude Code. # What is dsh-simple-cli
Read Moreclaude2dsh: Migrate Claude Code sessions, skills, and memory to DSH native sessions
`claude2dsh` is an MIT-licensed DSH plugin (version 0.2.0-rc.5 candidate) maintained by kirkchinese. It is used to migrate Claude Code's sessions, skills, global instructions, and project memories into DSH-native continuous sessions, and can also export DSH sessions back to Claude Code JSONL. Core commands include session import (returning counts for categories such as previewed, imported, and skipped), skill and global instruction migration, and converting project memories into DSH skill packs. All of these can be performed via a browser UI wizard started by `dsh web`. Installation requires Node.js >= 22.19.0, pnpm, and the dsh CLI. Regarding security design, the source Claude directory defaults to only...
Read Moreclaude-in-dsh: Delegate dsh web sessions to local Claude Code CLI
claude-in-dsh is a community plugin within the DSH plugin ecosystem, designed to leverage the local Claude Code CLI to drive specific sessions without replacing DSH's native web conversation interface. This plugin binds DSH sessions to the local Claude CLI, preserving DSH's native transcript display, tool cards, approval workflows, and command panel, thereby eliminating the tediousness of manually moving output. Core capabilities include mutually exclusive session engine switching, permission profile management (e.g., manual, auto, etc.), native interaction with approval and question cards, dynamic model switching, and subscription usage display. By utilizing an independent Broker hosting process, it ensures that DSH restarts or plugin hot updates do not interrupt running tasks and supports resume from interruption. Prerequisites include Node.js 20+ and pnp.
Read Moredsh-crew: Run Claude Code and Codex terminal panes side-by-side in dsh Web UI
`dsh-crew` is an MIT-licensed DSH plugin released by `dshworks`, designed to integrate Claude Code and Codex as real-time terminal panes into the DSH Web UI to enable human-machine split-screen collaboration. It allows users to intuitively monitor AI execution processes within a real PTY environment and supports inputting, interrupting, or broadcasting messages at any time. The plugin provides DSH agents with tools such as `crew_list`, `crew_seat`, `crew_send`, `crew_peek`, and `crew_dismiss`, enabling agents to manage member seating assignments, send tasks, read current screen content, and terminate processes. The `crew_send` function waits for the screen to stabilize before returning only new content, supporting...
Read Moredsh-claude-marketplace: Load Claude Code assets as DSH Skills
`ben7am1n/dsh-claude-marketplace` is a Claude Code compatible layer plugin for DeepSeek Harness (DSH), designed to enable DSH to directly reuse skills, commands, and agents from the project's `.claude` directory or the Claude Code marketplace, eliminating the need to maintain separate invocation methods. The plugin loads components from various sources—including local paths, GitHub, and URLs—via tool entry points such as `cc_scan` and `cc_marketplace_add`, registering them as native DSH skills with a `cc-` prefix. It is installed using the `dsh plugin` command and supports configuration of a cache directory.
Read Moredeepseek-harness-memory: DSH's Claude Code-style long-term memory plugin
deepseek-harness-memory is a long-term memory plugin for the DSH community, maintained by 2303572348 and licensed under the MIT License. The plugin is designed to provide cross-session persistent memory capabilities and stores data as plain markdown files (including YAML frontmatter), fully compatible with Claude Code memory formats. Key features include: supporting user-level and workspace-level memory storage, and automatically injecting memory indices into system prompts upon session start; providing a 'memory' tool for CRUD operations on memories and importing memories from Claude Code; and supporting the mapping of multiple checkouts to unified project memory via projectAliases. The plugin can be configured via cordis.patch.y...
Read Moredsh-peer-link: Native Point-to-Point Messaging for DSH Sessions
dsh-peer-link is a DSH plugin designed to enable point-to-point communication between DSH sessions and other agent sessions (such as Claude Code) on the same machine via Unix Socket. This plugin registers the DSH process as a discoverable peer, distinguishes specific sessions through short IDs, and supports injecting received messages into DSH sessions in the form of user bubbles. Core functionalities include `peer_send` (for sending messages to a specified peer or socket path) and `peer_list` (for viewing active peers, with support for directory filtering, sorting, and interactive operations). Installation requires using the `dsh plugin` command to add the plugin repository. If communication with Claude Code is needed, Claude Co
Read MoreDSH Crew: Assigning Tasks to DSH Agents in Claude Code / Codex
DSH Crew is a workflow plugin for DeepSeek Harness, maintained by ZSeven-W. It enables Claude Code, Codex, Antigravity, and Grok to dispatch tasks to a genuine DSH agent while displaying native sub-agent progress within the host. It supports flash/pro tier strategies, failure-based tier escalation, dispatch guards, and a task board; the DSH Web runtime worker is visible as a first-class session. It also provides visual description and image generation bridging for pure-text DSH models. The installation command is: `dsh plugin --profile web add @zseven-w/dsh-crew@latest`.
Read MoreIntegrate Codex and Claude Code into resumable DSH sub-agents using dsh-plugin-product-subagents
`dsh-plugin-product-subagents` is a DeepSeek Harness session and messaging plugin maintained by shaokeyibb (MIT License, npm version 0.3.1). It connects ACP CLIs such as Codex, Claude Code, and Cursor/CodeBuddy/Gemini/OpenCode to the dsh sub-agent channel, providing features including resumable subtask chats, remote session recovery, a declarative role library (general/code-review/explore/debug), and delegation permission ceilings. This article verifies the installation commands, six tools and security boundaries against the community directory page, GitHub README and official sub-agent documentation.
Read MoreImport Historical Sessions Such as Claude Code and Codex into DeepSeek Harness Using dsh-chat-import
dsh-chat-import is a DeepSeek Harness session and message plugin (MIT License, current version 0.5.1) maintained by Nwflower. It can read-only import historical conversations from 14 sources including Claude Code, Codex, ChatGPT, Cursor, Gemini, Reasonix and Kimi into resumable dsh sessions, and supports export, bundle backup and slash command handover. This article verifies the installation commands, usage and permission precautions based on the community directory page and GitHub repository.
Read MoreWrite Your First Agent Skill Using the Official template-skill
Template-skill is the starting skeleton of Skill in Anthropic's official repository anthropics/skills, with only one SKILL.md file containing name and description placeholders in the directory. Agent Skills became an open standard in December 2025, and the same file can be enabled by scanning directories in Cursor, Claude Code, and Codex CLI respectively. This article follows the official template and agentskills.io specifications to explain the frontmatter fields, directory structure, installation paths of each tool, and how to fill in the template to create the first triggerable Skill.
Read More# shadcn Skill: Let AI Search, Install and Combine shadcn/ui Components According to Project Configuration
shadcn is the official Agent Skill for shadcn/ui. After installation, it will read the project's components.json, inject the framework, aliases, installed components, icon library and underlying primitives via `shadcn info --json`, then search, install, debug and stitch together UIs according to the official combination rules. It was released alongside CLI v4 in March 2026, and can be added to tools like Cursor, Claude Code and Codex via `pnpm dlx skills add shadcn/ui`, preventing assistants from writing components that do not match the project based on memory.
Read MoreClaude Code v2.1.224: Multiple independent sessions can send messages directly to each other, say goodbye to copy-pasting between terminals
Anthropic launched cross-session messaging in Claude Code v2.1.224 (released on August 7, 2026): independently running sessions on macOS/Linux can discover each other via ListAgents and send each other plain text summaries via SendMessage, without manually copying context between terminals. The messages do not carry conversation history or files, and the recipient still needs to approve sensitive operations on their own; the communication uses local sockets on the same machine, and cross-machine support is limited to replies via Remote Control. This article introduces the mechanism, scenarios, usage and security boundaries based on official documentation.
Read MoreClaude Code Launches Default Auto Mode on August 14: AI Programming Agents Move from "Review Line-by-Line" to Autonomous Execution
Anthropic announced that starting August 14, newly created Claude Code sessions for Pro, Max and Team users will enable Auto Mode by default, replacing one-by-one permission pop-ups with a safety classifier. Official controlled experiments show that the classifier intercepts approximately 89% of dangerous commands, far higher than the approximately 14% interception rate of manual approval; PR output for Team users increased by about 25%. This article sorts out the Auto Mode mechanism, safety data, applicable scope of each subscription plan, and developer response suggestions.
Read More# migrate-to-codex: Official Migration Skill to One-Click Move Claude Code Configuration to Codex
OpenAI's official Skill `migrate-to-codex` can convert Claude Code's instructions, Skills, Hooks, MCPs and sub-Agents into Codex's standard directory structure, supports scan, plan, dry-run and verification, and outputs a migration report. This article introduces installation and activation, typical commands, as well as the complementary relationship and limitations between it and Codex /import.
Read MoreDetailed Explanation of canvas-design Skill: Incorporating Design Philosophy into AI to Generate PNG/PDF Static Visual Works
Canvas-design is a creative Agent Skill in Anthropic's official Skills repository. It adopts a two-step workflow of "first formulating design philosophy, then expressing it on the canvas, and finally conducting secondary refinement", and outputs .md manifesto and .png/.pdf visual finished products. This article verifies its core capabilities, canvas-fonts font resources, installation methods in Cursor and Claude Code, as well as typical usages such as posters and brand visuals and copyright considerations based on the official SKILL.md.
Read More# doc-coauthoring: Use a three-stage structured workflow to let Agents help you write clear technical specifications and decision documents
Doc-coauthoring is a documentation co-creation Skill in Anthropic's official Skills repository, targeting structured writing scenarios such as technical specifications, decision documents, RFCs, and PRDs. Instead of generating a full article at one time, it guides the Agent through three phases: context collection, section-by-section polishing, and reader testing. It first supplements background information through meta-questions and information dumping, then conducts brainstorming, screening, drafting and iterative revision by chapter, and finally uses a context-free Claude to simulate reader questions to identify blind spots in the document. This article introduces its core capabilities, installation methods in Cursor and Claude Code, as well as typical interaction examples and applicable boundaries for writing decision docs.
Read MoreAnthropic PDF Skill: Enabling AI Programming Assistants to Truly Handle All Scenarios of PDF Processing
This PDF is an Agent Skill maintained officially by Anthropic, which follows the Agent Skills open standard and can be used in Cursor, Claude Code and Codex. It covers high-frequency office automation scenarios including PDF reading and extraction, merging and splitting, rotating and watermarking, form filling, encryption and decryption, OCR and image extraction, etc., and includes tool selection guides for pypdf, pdfplumber, reportlab, qpdf, pdftotext and other tools. This article introduces the positioning, core capabilities, installation methods in Claude Code/Cursor/Codex, as well as official code examples and usage notes of the Skill.
Read More# Claude-api Skill: A Pocket Reference Manual for Claude API Developers
Anthropic's open-source claude-api Agent Skill packages the documentation of Messages API, Managed Agents and 8 programming language SDKs including Python, TypeScript and Go into progressively loadable references, which is built into Claude Code. This article introduces its positioning, core capabilities (model migration, streaming, tool calling, Prompt Caching, MCP/Agent), installation method and typical usage, helping developers avoid writing outdated Claude API code from memory.
Read MoreClaude Code 2.1.220 Defaultly Switches to Opus 5, Nested Subagent Depth Expanded to 3 Layers
On July 24, 2026, Anthropic released the 2.1.220 series updates for Claude Code: version 2.1.219 switched the default model to Claude Opus 5 (1M context window), and expanded the default nested subagent depth from 1 to 3; version 2.1.218 changed `/code-review` to run as a background subagent. An early August evaluation showed that the gap between Opus 5 and GPT-5.6-Sol on Terminal-Bench 2.1 was less than 0.5 percentage points, and the competition for AI programming agents has shifted to focusing on both models and orchestration layers. This article sorts out the version changes based on the official CHANGELOG, and provides configuration instructions for subagent depth limits, concurrency upper limits and other parameters.
Read MorePonytail: Let AI Agents Follow the YAGNI Principle to Reduce Over-Engineering and Token Waste
# Ponytail: August 2026 GitHub Trending Project Ponytail is a YAGNI (You Aren't Gonna Need It) agent skill suite designed specifically for coding agents. Through decision ladders and commands such as `/ponytail-review` and `/ponytail-audit`, it prioritizes reuse, standard libraries and native capabilities before writing code, curbing over-engineering and unnecessary dependencies. Official agentic benchmarks conducted on real FastAPI+React repositories show an average code reduction of approximately 54%, a token reduction of around 22%, a cost reduction of roughly 20%, while maintaining 100% security. It supports over 14 hosting environments including Claude Code Plugin and Cursor rule injection. This article introduces its principles,实测 data, installation methods and applicable boundaries.
Read MoreGraphify: Convert code bases, documents and SQL into queryable knowledge graphs to replace traditional vector RAG
Graphify, which has been consistently ranking on GitHub Trending in early August 2026, allows users to build locally queryable knowledge graphs from code, documents, SQL schemas, PDFs and more via the /graphify Skill. It uses tree-sitter AST for deterministic parsing at the code layer and does not rely on vector databases. The edges are labeled EXTRACTED/INFERRED for interpretability. This article introduces its differences compared to vector RAG, the steps to install and integrate it with Claude Code/Cursor, as well as its query, path and explain usages.
Read MoreNous Research Releases Open-Source Programming Model NousCoder-14B, Competing with Claude Code Moment
NousCoder-14B is an open-source competitive programming model developed by Nous Research based on Qwen3-14B and trained using verifiable reward reinforcement learning, achieving a 67.87% Pass@1 score on LiveCodeBench v6. This paper outlines its training stack (Atropos, DAPO, Modal), the differences in product form between it and Claude Code, and provides key points for local deployment with Transformers and Ollama. The 67.87% score is the official stated metric, and high scores on algorithm benchmarks do not equate to repository-level Agent capabilities; the open-source weights and complete RL pipeline are the long-term value of this release for developers.
Read MoreAnthropic Releases Opus 5: New Default for Programming Agents, Half the Price of Fable 5 with Comparable Capabilities
On July 24, 2026, Anthropic released Claude Opus 5, making it the default model for Claude Code. Its pricing is the same as that of Opus 4.8, at $5/$25 per million tokens. The official stated that its capabilities are close to those of Fable 5, but the cost is only about half. The model enables adaptive thinking by default, and the effort parameter can be adjusted across five levels. It topped evaluations including Frontier-Bench v0.1 and GDPval-AA, and achieved 96.0% on SWE-bench Verified. On the same day, GitHub Copilot integrated Opus 5, covering multiple terminals such as VS Code and CLI. This article sorts out the evaluation data, key points of API migration, and suggestions for selecting between Opus 5 and Fable 5.
Read MoreThe Two AI Coding Powerhouses in 2026: How Claude Code and Cursor Will Divide Labor
In 2026, a new consensus has formed in the developer community: Claude Code (terminal Agent + Claude Agent SDK) and Cursor (AI IDE) are often used in combination rather than being an either-or choice. Based on official documentation from Anthropic and Cursor, as well as public reviews from Faros.ai, FutureProofing.dev and other sources, this paper compares the differences between the two tools in terms of MCP integration, subagents orchestration, multi-model routing, Bugbot review and token efficiency, proposes a practical division of labor scheme such as "Cursor for daily coding, Claude Code for cross-file refactoring and CI automation", and explains that Claude Code can coexist as an extension of Cursor.
Read MoreClaude Opus 5 Released: 1M-Context Window Becomes Default Flagship Model of Claude Code
On July 24, 2026, Anthropic released Claude Opus 5, which features native 1 million-token context window, default thinking mode, and the same pricing as Opus 4.8. For Claude Code v2.1.219 and above, accounts such as Max/API will parse "Default" as Opus 5, while the default setting for Pro tier remains Sonnet 5. This article sorts out the specifications of Opus 5, Fast Mode and Effort levels, as well as the toolchain updates of Claude Code in July including Subagent background running, nested delegation, /code-review background review and /doctor self-check, and provides step-by-step guides for version checking and model switching.
Read MoreGraphify: An Agent Skill for Translating Code Repositories, Documents and Schemas into Queryable Knowledge Graphs
Graphify, which trended on GitHub Trending in early August 2026, constructs queryable knowledge graphs from code, documents, SQL schemas, PDFs and more via local AST powered by tree-sitter and Leiden community detection, and integrates with Agent tools including Claude Code, Cursor and Codex through the `/graphify Skill`. This article verifies its Trending background and official capability boundaries, makes a comparison with vector RAG, and provides hands-on steps for installation via uv, Skill registration and querying via query/path/explain, to assist with context engineering选型 in large codebase scenarios.
Read MoreAgent Orchestrator: A Parallel Orchestration Platform for IDE to Manage 23 Terminal Coding Agents
Released by Composio as an open-source Agent Orchestrator (AO) in February 2026, this is an orchestration layer for parallel-coding agents: it supports 23 types of terminal agents including Claude Code, Codex and Cursor, adopts Git Worktree isolation by default, and automatically routes CI failures and review comments back to the corresponding sessions. It has gained over 8,700 GitHub stars so far. This paper outlines its workflow, plugin architecture and automatic feedback loop, and explains its complementary relationship with single-agent IDEs.
Read MoreOrca: An ADE Running Five Coding Agents Simultaneously, A New Parallel Orchestration Paradigm with Over 20,000 Stars on GitHub
Orca, an open-source project by Stably AI, was a trending ADE (Agent Development Environment) on GitHub Trending in July 2026. Built on Git Worktree, it provides isolated parallel runtime environments for over 30 CLI Agents including Claude Code, Codex, and OpenCode. It supports one-click Worktree creation from GitHub/Linear tasks, UI context injection via Design Mode point-and-click, as well as SSH remote Worktree. Its GitHub stars exceeded 20,000 in July and continued to rise. This article sorts out the differences between ADE and traditional IDEs, core mechanisms, installation methods and applicable scenarios, to help evaluate multi-agent parallel development solutions.
Read MoreOrca Going Viral on GitHub: Run Claude Code, Codex and Cursor in Parallel in ADE
In July 2026, the open-source project Orca (stablyai/orca) topped the GitHub Trending weekly chart, with its star count exceeding 30,000. Positioned as an Agent Development Environment (ADE) rather than a traditional IDE, it allows running terminal Agents such as Claude Code, Codex, and Cursor CLI in parallel on the same interface. Each Agent uses an independent git worktree to isolate changes, and supports prompt fan-out comparison, Diff annotation review, GitHub/Linear integration, and orchestration via the Orca CLI. Based on official repository and public Trending analysis, this paper sorts out Orca's core mechanisms, functional boundaries, and selection comparisons with herdr and IDE-built-in Agents, providing reference for developers in multi-Agent parallel coding scenarios.
Read More