跨会话记忆桥：让 AI 助理跨会话接续项目上下文
跨会话记忆桥（@user_41b104fd/context-bridge-pro，v1.0.3）是 SkillHub 社区 ai-agent 类 Skill，无需 API Key。它作为主动 Agent 在后台记录多项目的进度、决策与待办，支持模糊匹配、别名与自动发现新项目，便于跨会话、跨设备接着聊。安装需按 skillhub.md 使用 skillhub CLI，并用 --dir 指向 Cursor、Claude Code 或 Codex 等对应的 skills 目录。装上后可用「继续上次」或官方列出的中英文触发词接续上下文，适合多项目并行与长周期任务；存储与同步细节以 Skill 实际行为为准。
Read MoreContextengineering：用 Skill 把 Agent 上下文当成工程问题
Contextengineering 是 SkillHub 社区 Skill（@user_3c6cb52e/context-engineering-lr-314c，v1.0.0），无需 API Key，帮助 Agent 按持久规则、规格文档、源码、错误输出与对话历史五层组织上下文。文章说明适用场景、核心层次与 CLAUDE.md 规则文件思路，并给出 skillhub 安装命令及 Cursor、Claude Code、Codex 的 skills 目录。注明发布包 SKILL.md 存在截断，安装时需核对命名空间。
Read More上下文工程 Agent 技能：为 Agent 任务补齐上下文工程指引
本文介绍 SkillHub 社区技能「上下文工程 Agent 技能」（@user_3c6cb52e/context-engineering-collection-lr-di68，v1.0.0，分类 ai-agent）。它基于通用 SKILL.md，面向需要为 Agent 固化上下文工程约定的开发者。文章依据 SkillHub 页面与安装文档说明 skillhub CLI 安装、Cursor/Claude Code/Codex 的 skills 目录、requires_api_key=false 等可核对事实；因公开页未展示技能正文，核心能力与示例以安装后本地 SKILL.md 为准。
Read MoreEvaluation Skill: Building a Reproducible Evaluation Framework for Agent Systems
evaluation, from Agent Skills for Context Engineering, guides designing a reproducible evaluation system for agent systems: outcome-oriented, weighted multi-dimensional rubrics, deterministic checks prioritized over LLM judgment, and coverage of test-set stratification, continuous evaluation, and production monitoring. The article explains when to enable it, how responsibilities are divided with skills such as advanced-evaluation, and directory-level installation methods in Cursor, Claude Code, and Codex, excerpting official pseudocode examples and common evaluation pitfalls.
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 Moremulti-agent-patterns: Multi-Agent Collaboration Architectures and Context Isolation
multi-agent-patterns, from Agent-Skills-for-Context-Engineering, targets multi-agent system designs that require context isolation, supervisor or swarm coordination, explicit handoffs, and parallel execution. The skill emphasizes that sub-agents should primarily be used for context isolation rather than role-play, and summarizes three patterns—Supervisor/Orchestrator, Peer-to-Peer/Swarm, and Hierarchical—along with the tradeoffs involved in selecting among them. It covers the telephone-game problem and direct forwarding via forward_message, includes cautions about token consumption relative to single-agent designs, and clarifies boundaries with adjacent skills such as project-development and hosted-agents. It is suitable for developers making architecture decisions and running baseline evaluations in production-grade, multi-component agent systems.
Read MoreSelf-Improvement Loops: When the Harness Itself Becomes the Optimization Target
self-improvement-loops is a Skill newly introduced in Agent Skills for Context Engineering 2.4.0, targeting scenarios where the harness, workflow, or optimizer itself becomes the optimization target. The article explains its optimization ladder, out-of-loop evaluation invariants, empirical acceptance gating based on held-in/held-out data, filesystem-backed experience archives, and failure-driven three-stage self-editing, and provides directory-based installation instructions and official examples for Cursor, Claude Code, and Codex, helping it coordinate with adjacent Skills such as harness-engineering and evaluation.
Read MoreHarness Engineering: Designing Runtime Frameworks for Autonomous Agents
harness-engineering is a skill within Agent Skills for Context Engineering, used to design the runtime harness for autonomous agents: it distinguishes locked, editable, append-only, and human-controlled surfaces; establishes tight feedback loops and persistent THREAD logs; pairs novelty gating, pruning, rollback, and PR preparation; and clarifies human approval boundaries for merge and deployment. Based on the official SKILL.md, the article outlines the Autoresearch and Research-To-Skill loops, key anti-metric-gaming considerations, and the directory layout, explaining when to enable it and how it divides responsibilities with adjacent skills such as evaluation.
Read Morememory-systems: A Design Guide for Agent Persistent Memory and Retrieval Architectures
memory-systems is the Memory System Design guide in the Agent Skills for Context Engineering collection, covering cross-session semantic memory, entity consistency, temporal validity, graph/vector and hybrid retrieval, and memory consolidation. The article explains its positioning, activation boundaries, memory hierarchy and framework selection table, and principles for consolidation and JIT loading. It also provides official examples using Mem0, temporal graphs, and Cognee, along with loading methods via Claude Code and npx skills, emphasizing starting with a shallow design, validating it with LoCoMo/LongMemEval, and then deepening the architecture.
Read Moredsh-context-engineering: Context Engineering Planning Based on a Five-Tier Hierarchy
`dsh-context-engineering` is a DSH plugin designed to optimize agent input information through a proactive curation mechanism. Inspired by the agent-skills project, it focuses on context engineering, with its core capability being the structured organization of input information into a five-tier hierarchy, ensuring that the agent receives filtered and planned information before processing tasks. The plugin runs with current process permissions and is suitable for development scenarios that require strict context control. Currently, installation commands and examples are unavailable due to a README fetch failure.
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 MoreModel Isn't the Decisive Factor: The Harness Battle for CLI Coding Agents in 2026
In mid-2026, the focus of the CLI coding Agent community shifted from foundation models to Harness—system prompts, retry logic, context compression, and sub-agent orchestration. On Terminal-Bench 2.1, Claude Code and Codex CLI were locked in a tight race with an accuracy rate of around 83%, and the same model could achieve scores differing by several percentage points under different Harness configurations. This paper sorts out the concept of Harness, compares four development routes including Claude Code, Codex CLI, OpenCode and Copilot CLI, and explains why context compression has become an invisible battlefield, to help developers select appropriate solutions according to their scenarios.
Read More