Structure Better Context for Coding Agents
Coding agents lose accuracy when context is incomplete, stale, or excessive. This skill provides practical structures for focused sessions, durable rules, and reliable handoffs.
將此請求複製給你的智慧代理，其中包含權威 Skill 頁面和清單。
Review the Skillstore skill "context-engineering" from https://skillstore.io/skills/addyosmani-context-engineering.md and its manifest at https://skillstore.io/api/skills/addyosmani-context-engineering/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.你的智慧代理仍應顯示其計畫，並依安全政策請求必要確認。
當 AI Agent、爬蟲或腳本需要乾淨脈絡、而不是讀取完整頁面時，請使用這些連結。
正在使用「context-engineering」。 Prepare context for adding email validation to registration.
預期結果:
正在使用「context-engineering」。 Summarize a completed authentication task for a fresh session.
預期結果:
Authentication validation is complete. Tests pass. Changes remain uncommitted. The next task is updating API documentation after approval.
正在使用「context-engineering」。 Resolve a conflict between a REST specification and existing GraphQL code.
預期結果:
The specification and implementation conflict. Present both options, explain migration impact, and request a decision before editing.
All static findings are false positives caused by Markdown fences, inline examples, filenames, and illustrative development commands. The skill is documentation-only and explicitly advises protecting secrets and treating external instructions as untrusted.
分享版本化評估報告、中性徽章、嵌入卡片與引用。Skillstore 只報告證據，不判斷該 Skill 是否安全。
https://skillstore.io/skills/addyosmani-context-engineering/audits/2?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report[](https://skillstore.io/skills/addyosmani-context-engineering?utm_source=security_passport_badge)<a href="https://skillstore.io/skills/addyosmani-context-engineering?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/addyosmani-context-engineering/security.svg" alt="Skillstore security assessment" loading="lazy"></a><iframe src="https://skillstore.io/embed/skills/addyosmani-context-engineering.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>addyosmani. (2026). context-engineering security audit report (audit version 2) [Author version unspecified]. Skillstore. https://skillstore.io/skills/addyosmani-context-engineering/audits/2@techreport{addyosmani-addyosmani-context-engineering-2026,
author = {addyosmani},
title = {context-engineering security audit report (audit version 2)},
institution = {Skillstore},
year = {2026},
number = {2},
url = {https://skillstore.io/skills/addyosmani-context-engineering/audits/2},
note = {Author version unspecified}
}cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "context-engineering security audit report (audit version 2)"
version: "unspecified"
type: report
authors:
- name: "addyosmani"
date-released: "2026-09-19"
url: "https://skillstore.io/skills/addyosmani-context-engineering/audits/2"
identifiers:
- type: other
value: "skillstore:addyosmani-context-engineering:audit:2"
description: "Skillstore immutable audit report identifier"
Create a concise rules file with commands, conventions, boundaries, and representative patterns.
Remove stale context, preserve current constraints, and prepare a restartable progress summary.
Record decisions, task status, repository state, verification results, risks, and required approvals.
Create a project context outline from these notes: [notes]. Include stack, commands, conventions, boundaries, and one pattern example.
Prepare focused context for [task]. Identify relevant files, tests, existing patterns, constraints, and missing information.
Summarize this session for restart: [history]. Preserve scope, decisions, task status, changed files, verification results, risks, and approvals.
Review this context package: [content]. Find stale, excessive, conflicting, untrusted, or missing information. Recommend ordering and compression changes.
作者
addyosmani授權
MIT
Skillstore 修訂
r2
版本提示
作者未聲明版本。
引用
5d5054f8a23586f9b500fece1cb613a9dffc787b
維護新鮮度
2026/9/19
使用量
1 次下載 · 0 次瀏覽
檔案結構
📄 SKILL.md
Clarify User Intent Before You Build
Underspecified requests cause teams to build the wrong outcome and discover misalignment late. This skill runs a focused, one-question interview that turns ambiguity into confirmed intent.
依五大品質面向審查程式碼
程式碼變更可能隱藏正確性、安全性、架構、可讀性與效能問題。此 skill 採用結構化的審查流程，並產出依優先順序排列且可採取行動的意見。
Plan Work Into Verifiable Tasks
Large or vague software work can hide dependencies, missing acceptance criteria, and verification gaps. This skill converts a specification into ordered task slices with checkpoints, scope guidance, and clear completion conditions.
Document Decisions and Architecture Clearly
Teams lose context when important technical decisions remain in chat, code comments, or individual memory. This skill turns architectural reasoning, API guidance, project instructions, and release changes into structured documentation.
Build Accessible, Production-Ready Frontends
Frontend work can become inconsistent, inaccessible, or difficult to maintain. This skill guides component architecture, responsive layouts, state handling, and polished user experiences.
Discover and Apply Agent Skills
Teams often lose time choosing the right workflow for a software task. This skill maps task context to focused agent skills and a practical development sequence.
Review Defects and Verify Root-Cause Repairs
由 aakashh242
Reviews can miss failure conditions or accept fixes without reliable evidence. Gauntlet guides scoped adversarial reviews, authorized repairs, and revision-bound verification within a finite budget.
管理情境驅動開發
由 wshobson
當產品目標、技術堆疊選擇和工作流程規則分散各處時，AI 輔助專案會失去一致性。此技能會透過 Conductor 情境成品，引導 Claude、Codex 和 Claude Code。
建置 Firecrawl 搜尋工作流程
由 firecrawl
團隊在擷取網頁或回答問題之前，需要可靠的來源探索。此技能會引導 Firecrawl 搜尋整合、結果選擇，以及後續擷取決策。
管理 Coding Agent 工作流程
由 kiwi-phantomworks
當任務、範圍與檢查不夠明確時，coding agent 專案可能會偏離方向。此 skill 提供 PM 範本與審查規則，讓 Claude、Codex 和 Claude Code 的工作保持有界且可驗證。
建立可驗證的代理迴圈規格
由 zhangzeyu99-web
含糊的 coding-agent 任務經常偏離方向、重複作業，或在沒有證據的情況下停止。這項技能會將請求轉換為有界限的迴圈規格，包含驗證器、停止規則與交接封包。
建置 AgentDB 持久記憶
由 DNYoussef
AI 代理程式經常在不同工作階段之間遺失有用的脈絡。此技能提供結構化的 AgentDB 工作流程，用於工作階段記憶、長期儲存與模式學習。