資料來源#
- Agent Data Injection Attacks are Realistic Threats to AI Agents
- Agentic coding and persistent returns to expertise
- Anthropic's Boris Cherny: Why Coding Is Solved, and What Comes Next
- Auto mode for Claude Code
- Best Practices for Claude Code
- Full Walkthrough: Workflow for AI Coding — Matt Pocock
- How Anthropic's product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code)
- Introducing Claude Opus 4.7
- Introducing Claude Sonnet 5
- The Founder's Playbook: Building an AI-Native Startup
- Zero Trust for AI Agents
摘要#
Anthropic 的代理式編碼產品,由 Boris Cherny 於 2024 年底在內部孵化器(Anthropic Labs)中建立。最初是輸入提示取代工具;搭配 Sonnet 3.5 後轉向代理式模式;在 Opus 4(2025 年 5 月)帶來轉折前,約有 6 個月處於 pre-PMF 狀態;之後 Opus 4.5、4.6 與 4.7 又各自帶來轉折。截至 2026 年 5 月,它已橫跨 CLI、桌面、網頁、行動裝置與 IDE(VS Code、JetBrains)提供服務——同一個產品介面,不同的進入點。
程式碼庫#
根據 Boris Cherny 的說法(2026 年 3 月因發布流程 PR 中的人為疏失而公開洩漏——之後已強化,並非惡意洩漏),其實作並不特別:TypeScript + React。這是刻意的選擇:在語言涵蓋度仍很重要的 2024 年底,這對模型而言「非常符合分布」。
Boris 的主張是:「一年後只剩 100 行程式碼」——關於這個方向而非字面主張,請參見 Harness Shrinkage as Models Improve。
介面形式#
- CLI — 主要介面,最先取得新功能。「所有工具中最強大的工具」(Cat Wu)。
- 桌面應用程式 — 提供前端工作的預覽窗格整合;更適合非技術使用者。
- IDE 擴充功能 — VS Code、JetBrains;根據 Boris 的受眾調查,其使用量占比明顯低於 CLI。
- 行動裝置 + 網頁 — 用於 AFK 啟動任務;Boris 每日使用的主要工具。
- 姊妹產品:Cowork — 使用相同原語,但輸出非程式碼內容。
顯著功能#
/loop— 由 Claude 排程的 cron 工作;Agent Loop Pattern 的主要原語/powerup— 帶你瀏覽現有的 100 多項功能- 子代理 — 彼此隔離 token 的內容視窗,並回報摘要;參見 Context Window Smart Zone
- Skills — 儲存在儲存庫中的 markdown 檔案,Claude 可按需載入;參見 Deep Modules for Agents 中的 push/pull
- Auto mode — 以分類器為基礎的權限閘控;參見 Claude Code Auto Mode
- Routines — 伺服器端的
/loop;即使關閉筆電仍會持續運作 - 程式碼審查 — 斜線指令 + 多代理審查模式;根據 Cat Wu 的說法,在 Opus 4.5/4.6 左右變得可靠
- Claude Code Security — 有限測試版(2026 年 5 月);掃描程式碼庫尋找安全漏洞,並提出供人類審查的精準修補方案。能發現傳統方法遺漏的問題。定位為從 MVP→Launch 的轉換階段(參見 AI-Native Startup Lifecycle);明確不是人類審查者或合格合規審查的替代品
安全態勢(Zero Trust 參考實作)#
Zero Trust for AI Agents 幾乎以 Claude Code 作為每個控制層級的運作範例——其安全原語對應該框架的八個控制領域:
- 預設拒絕權限 + 寫入存取限制,僅限於專案目錄 → Least Agency/Blast Radius (Agentic)
- Sandboxing(檔案系統 + 網路隔離、作業系統層級強制執行)→ 資源邊界/爆炸半徑遏制
- MCP 連線使用 OAuth 2.0 自動重新整理、作業系統憑證儲存區、
apiKeyHelper、工作階段範圍的「ask」權限 → Agent Identity and Authentication - 指令封鎖清單(預設封鎖 curl/wget)、網頁內容使用隔離的內容視窗、指令注入偵測、網路請求核准 → Agentic Prompt Injection 防禦
- 預設工作階段隔離 +
cleanupPeriodDays+ 檢查點/倒帶 → Memory and Context Poisoning 防護 - Managed settings/
allowManagedPermissionRulesOnly/伺服器管理(MDM)的設定 → 組織層級治理,使用者無法覆寫 - Hooks(PreToolUse 參數驗證、ConfigChange 稽核)+ 使用每個工作階段
session.id的 OpenTelemetry 遙測 → 可追溯性與工具存取控制
已證實的攻擊面(Agent Data Injection)#
Choi 等人(arXiv 2607.05120,2026 年 7 月;已負責任地揭露並獲 Anthropic 確認)展示了針對 Claude Code 的可運作 Agent Data Injection 漏洞利用——這是該儲存庫首個經實證的 RCE:
- 透過來源注入執行遠端程式碼 — GitHub issue 留言偽造含有冒充維護者作者/角色的留言物件(將機率式分隔符注入
gh/GitHub-MCP JSON),當使用者要求 Claude Code「套用維護者的修正」時,誘使其執行攻擊者指令(pip install starship-auth)。 - 透過工具呼叫注入發動供應鏈攻擊 — PR 描述注入偽造的
<function_calls>/<function_results>標籤(Claude Code 自身的工具呼叫區塊分隔符),捏造無害的read_pr_commit結果;Claude Code「審查」偽造的無害差異,並合併真實提交為 XSS payload 的 PR。 - Slack 來源注入 — 透過 Slack MCP 伺服器,普通成員的訊息偽造頻道管理員訊息區塊,Claude Code 在摘要時將其歸因於管理員。
bash/merge 確認對話框並未阻止這些攻擊:因為代理自身顯示的推理建立在誤解之上,確認描述(「依維護者建議」)反而強化了攻擊者的虛假敘事。根本原因不是 Claude Code 的實作錯誤,而是共通的 IPI 弱點——未將受信任資料與不受信任資料隔離——因此也會同樣影響 Codex 與 Gemini CLI(Google DeepMind)。參見 Agent Data Injection (ADI)。
重要使用者(依來源)#
- Boris Cherny — 100% 的自身程式碼由 Claude Code 完成,單日提交 150 個 PR
- Cat Wu — 主要 IC 工具,推動功能交付速度
- Fiona Fung — 領導工程 + 產品;「我們用 Claude Code 建立 Claude Code」
- Andrej Karpathy — 重度使用者(「cloud code/codex/open claw」);將這套紀律描述為 agentic engineering
- Matt Pocock — 工作坊等級的工作流程;將 Skills 調整為 grill-me + Ralph loop 模式
- Thariq Shihipar — Claude Code 工程師;「HTML 是新的 markdown」工作流程(參見 HTML as the New Markdown)
- Anthropic Applied AI 團隊 — 僅次於工程團隊的內部 token 第二大使用者
鄰近/相依概念#
- Claude Code Best Practices — 官方最佳實務文件
- Claude Code Auto Mode — 權限分類器
- Claude Opus 4.7 — 目前模型
- Claude Sonnet 5 — Claude Code 自發布起即可使用的模型(2026 年 7 月);較便宜的代理式選項,其 effort levels 能讓它在部分任務上接近 Opus-4.8 的品質
- Mythos Model — 內部使用的預覽模型
- Agent Loop Pattern — loop 原語
- Loop Engineering — Claude Code 提供全部五種 loop 原語(Osmani 的剖析):
/loop+/goal+ cron/hooks(自動化)、git worktree/isolation: worktree、Skills、MCP + plugins、.claude/agents/子代理 - Harness Shrinkage as Models Improve — Boris 關於發展方向的主張
- Engineer PM Convergence — 建立它的團隊體現了這項趨勢
- Anthropic Labs — Claude Code 所出自的孵化器(2024 年底)
- Claude Design — 建立在 Claude Code 的 Agent SDK + 現有 Claude Code skill 之上;Claude Design 的「交接給 Claude Code」會將設計送回此處投入生產
- AI Native Product Cadence — Cat Wu 對團隊如何交付的闡述
- Agentic Misalignment (AM) — agent-mode + skip-permissions 部署會暴露 AM 威脅面;與無人值守的
/loop使用者相關 - Claude's Constitution / Model Spec — 形塑 Claude Code 性格與拒絕行為的價值觀
- AI Employee Framing — Claude Code 是同一產品問題中面向工程師工具的一側,而 HBR 研究的是 HR/財務情境
- HTML as the New Markdown — 團隊內部的 Claude Code 使用模式:以 HTML 產物作為面向人類的溝通媒介
- Disposable Micro-Apps/Living Design System — Thariq 的其他 Claude Code 工作流程
- Compute Allocator — Claude Code 日益服務的使用者角色:決定哪些事情值得投入算力,而不是輸入程式碼
- AI-Native Startup Lifecycle — Claude Code 作為四個創辦人階段中的主要 MVP 建置工具
- Agentic Technical Debt — 未持續維護 CLAUDE.md 內容時,使用 Claude Code 特有的失敗模式
- Zero-Friction Scope Creep — Claude Code 每項功能的時間成本低所造成的失敗模式
- Founder as Agent Orchestrator — Claude Code 為非工程背景創辦人促成的創辦人角色轉變
- Compounding Data Moat — Claude Code 建立垂直領域專屬測試套件與整合,作為護城河組件
- Problem-Solution Fit Discipline — Claude Code 在 Idea 階段的角色受限於作為討論道具的輕量原型——明確不是驗證;這套紀律讓工具維持在支援證據的角色
- Verification as the New Bottleneck — Fiona Fung:在 Claude Code 團隊,編碼不再是緩慢的部分;驗證/審查才是
- Code as Source of Truth — Claude Code 團隊的知識分享規範:程式碼 + 已簽入的 Skills 是真實來源,透過 Claude 完成入職
- Returns to Expertise in Agentic Coding/Planning / Execution Division of Labor/Agentic Coding Work-Composition Shift — Anthropic 對 Claude Code 使用情況進行的 40 萬工作階段經濟研究(2025 年 10 月至 2026 年 4 月)的三項發現:領域專業(而非編碼技能)會放大工具效益;人類作出約 70% 的規劃決策/Claude 作出約 80% 的執行決策;隨著價值上升約 27%,使用情況從除錯轉向端到端代理式工作
- Conversation-to-Delegation Shift — AEI 的 Cadences 報告發現,即使控制模型因素,Claude Code 仍比聊天/Cowork 帶來更高的 AI 自主性(一篇部落格 = 13 輪聊天,而 1 個 Code prompt;54% 的 Code 工作階段使用 Opus,而聊天僅 10%)——「產品比模型更重要」
- Vibe Coding vs. Agentic Engineering — Claude Code 作為這套紀律的典型介面
- Zero Trust for AI Agents — 將 Claude Code 作為所有控制層級參考實作的安全框架
- Least Agency/Blast Radius (Agentic)/Agent Identity and Authentication/Agentic Prompt Injection/Memory and Context Poisoning — Claude Code 原語實作的具體 Zero Trust 控制(見上方安全態勢)
- Agent Data Injection (ADI) — 針對 Claude Code 已展示的 RCE/供應鏈/Slack 來源漏洞利用(見上方攻擊面章節);確認對話框強化了攻擊者偽造的敘事
資料來源#
- Anthropic's Boris Cherny: Why Coding Is Solved, and What Comes Next
- How Anthropic's product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code)
- Best Practices for Claude Code
- Auto mode for Claude Code
- Introducing Claude Opus 4.7
- Full Walkthrough: Workflow for AI Coding — Matt Pocock
- The Founder's Playbook: Building an AI-Native Startup
- Agentic coding and persistent returns to expertise — Anthropic Economic Research,2026 年 6 月;40 萬工作階段的使用研究
- Introducing Claude Sonnet 5 — Sonnet 5 自發布起即可在 Claude Code 中使用(2026 年 7 月)
- Agent Data Injection Attacks are Realistic Threats to AI Agents — Choi 等人,arXiv 2607.05120;展示針對 Claude Code 的 ADI 漏洞利用(透過來源注入執行 RCE、透過工具呼叫注入發動供應鏈攻擊、Slack 來源注入)
Cited by 94
- Anthropic×8
2025 December — acquired Bun, the JavaScript runtime Claude Code is built on; Jarred Sumner and the…
- Boris Cherny×6
Creator and tech lead of Claude Code at Anthropic. Engineer-by-background, author of Programming…
- Cowork×6
How does Cowork's harness compare to Claude Code's? Both surface skills, MCP, sub-agents — but the…
- Opinions on Using AI Tools & the Future of the Software Engineering Role×5
"Coding is solved (for me)." Boris writes 100% of his code via Claude Code, has logged 150-PR days,…
- Learning to Co-Work with AI: A Software Engineer's Field Guide×5
Cross-disciplinary range matters more than vertical depth. Cat reports every functional role on the…
- Bun×4
This is not in tension with "Claude Code has shipped the Rust port since v2.1.181": Claude Code…
- Claude Design×4
Handoff to Claude Code — push a design into production without re-exporting files and re-typing the…
- MCP and Computer Use×4
Created at Anthropic Labs (late 2024) alongside Claude Code and the desktop app by Boris's founding…
- Orchestration Sets Token Economics×4
Does the effect survive against a competent third-party baseline rather than a vendor's own frozen…
- Anthropic Labs×3
Per Anthropic's entity page and Boris Cherny: a first incarnation of the Labs incubator formed in…
- Build for the Next Model×3
The over-shoot he warns about — "too AGI-pilled for the moment." Ambrosino names the failure mode…
- Cat Wu×3
Head of Product for Claude Code and Cowork at Anthropic. Engineer for many years before a brief VC…
- Claude Character as Product×3
This is the rare-trusted-evaluator pattern: Cat says "there's a handful of people who are much…
- Dynamic Workflows: An Algebra for Agents×3
Most of the apparent contradiction dissolves against Sumner's own account, and should not be staged…
- Memory and Context Poisoning×3
Everything above is threat taxonomy from a defense framework. bad memory (University of Washington…
- Open-Ended Discovery Harnesses×3
Claude Code — the runtime for both SwarmResearch and CORAL; the harness is implemented entirely as…
- OpenHands×3
OpenHands is an open-source coding-agent platform, and also the company that maintains and sells…
- Orchestration-Plan Simulation×3
The claim that makes it usable is the sim-to-real one: simulated scores correlate with real Claude…
- Shared Harness, Differentiated Surfaces×3
Anthropic answered two: Claude Code for work whose output is code, Cowork for work whose output…
- Agent Identity and Authentication×2
Identity is the prerequisite for Blast Radius containment (identity-based isolation: services…
- Agentic Coding Work-Composition Shift×2
The longitudinal finding of Anthropic's 400K-session study: over just seven months (Oct 2025 → Apr…
- Agentic Honesty & Diligence×2
These are exactly the failure modes that make autonomous agentic coding risky: when a model writes…
- Agentic Misalignment (AM)×2
This describes Cowork, Claude Code in agent mode (especially --dangerously-skip-permissions),…
- AI as Primary Author×2
Claude Code — agent mode (apply-changes-directly) is named as a primary driver of the 20%→60%…
- AI-Native Startup Lifecycle×2
the founders playbook building an ai native startup (Anthropic, May 2026). 36-page ebook organized…
- AI Usage Cadences×2
> Evidence note. empirical — privacy-preserving classifiers over continuously-sampled Claude.ai,…
- Anthropic Economic Index×2
The Anthropic Economic Index (AEI) is Anthropic's ongoing economic-research program studying how AI…
- Claude Sonnet 5×2
API model id claude-sonnet-5; the default model for Free and Pro plans, and available to Max, Team,…
- Codex×2
Claude Code — the Anthropic-side peer harness Codex is compared against (same five loop primitives,…
- Cursor×2
Authorship telemetry — Faros attributes the 20% → 60% rise in AI-code acceptance substantially to…
- Engineer PM Convergence×2
Both Boris Cherny (Sequoia AI Ascent 2026) and Cat Wu (Lenny's Podcast, April 2026) report the same…
- Evals as Product Spec×2
The Claude Code team at lunchtime vibe-checks — feedback like "this model isn't testing itself…
- FastContext×2
FastContext is the open, published counterpart to the proprietary subagent mechanisms inside Claude…
- Fiona Fung×2
Leads engineering and product for Claude Code and Cowork at Anthropic; previously built and led…
- Harness Build-vs-Buy×2
Claude Code — conspicuously absent from the comparison, being closed-source; the shrinkage claims…
- Kimi (Moonshot AI)×2
K3's evaluation footnotes are the most granular harness disclosure in this corpus, and they are…
- Loop Engineering×2
Loop engineering is replacing yourself as the person who prompts the agent — you design the system…
- OpenAI×2
On agent orchestration, Symphony/Codex (OpenAI) and Claude Code (Anthropic) are the two reference…
- Planning / Execution Division of Labor×2
Anthropic's 400K-session study supplies the empirical shape of human–agent collaboration in agentic…
- Repository Exploration Subagent×2
This is the open, published mirror of the proprietary "subagent" features shipping in Claude Code,…
- Returns to Expertise in Agentic Coding×2
The headline finding of Anthropic's economic-research report Agentic coding and persistent returns…
- Thariq Shihipar×2
Engineer on the Claude Code team at Anthropic. Source of the "HTML is the new markdown" thesis (see…
- Thinking Machines Lab×2
Their harness-dissolves-into-model stance is the same shape as Harness Shrinkage As Models Improve…
- Unknowns as the Agentic Bottleneck×2
The launch video for Fable was edited entirely by Claude Code, in a domain Thariq says he is "by no…
- Write-Then-Trusted×2
Claude Code / Codex / Google Deepmind — the affected agent products; the .claude hook-configuration…
- Zero-Friction Scope Creep×2
A failure mode identified in the founders playbook building an ai native startup: the traditional…
- Agent Data Injection (ADI)
Claude Code — the agent the RCE and supply-chain PoCs were demonstrated against; also the…
- Agent Harness Engineering
Legibility becomes the thing you tune per surface, and it has a cost. Sub-agent transcripts are…
- Agent Loop Pattern
Used inside Claude Code and Cowork. Mechanism: agent calls cron (via tool) to schedule a job at a…
- Agent-Native Infrastructure
Claude Code — the agent that consumes copy-paste skills and drives computer-use actuators
- Agent Review Comment Resolution
> Evidence note. empirical, confirmed on full read, with four qualifications that travel with every…
- Agentic Technical Debt
Debt that *compounds* (not just accumulates) because each agentic-coding session re-derives architectural decisions wit…
- AI Accelerating AI Development
Claude writes most of Anthropic's code. As of May 2026, >80% of merged code is Claude-authored, up…
- AI Native Product Cadence
Does the cadence scale beyond ~100 people? Anthropic itself is bigger (~30-40 PMs alone), but the…
- AI-Native Product Org Bottlenecks
For a small Claude Code-style team, the visible bottleneck is taste. The team can move quickly…
- Andrej Karpathy
Claude Code — names "cloud code / codex / open claw" as the agentic-coding surfaces he lives in
- Building Is Cheap, Arguing Is Expensive
Fiona Fung's rule for technical debates once agentic coding makes generation nearly free: "in…
- Capability Gating Is Not Authorization
Permissive-framework agents (Claude Code, Codex CLI, Gemini CLI) show substantially higher…
- Claude Code Best Practices
Anthropic's guide to effective Claude Code usage: context management, verification-driven development, explore→plan→cod…
- Claude Fable 5
Claude Code — the agentic runtime Mythos-class coding gains flow through
- Claude Opus 4.7
GA frontier model from Anthropic; direct upgrade to 4.6 at same price; literal instruction following, 1.0–1.35× tokeniz…
- Compounding Data Moat
Claude Code / Cowork / Anthropic — Skills, MCP integrations, and APIs are the surfaces this moat is…
- Compounding Loop Optimization
The handoff to Claude Code feature exists because the team kept re-typing, across tools, all the…
- Compute Allocator
Claude Code — the tool through which the allocator spends compute; increasingly serves "decide what…
- Cost-per-Task Over Cost-per-Token
Every source above answers which model or harness is cheaper. Simon Willison (2026-07-03,…
- Design by Selection
Not for shipping production software. "If you're shipping production software, stick with Claude…
- Disposable Micro-Apps
Claude Code — the product used to generate the throwaway UIs on demand
- Where Does Agent Harness Work Remain Durable as Models Improve?
Harness Shrinkage As Models Improve gives the negative space. Early Claude Code needed aggressive…
- Emergent
Indian AI coding / app-builder startup ("basically getting an engineering team in a box," per…
- Firm AI-Spend Intensity and Headcount Growth
US software-development postings +15% since Claude Code's launch (the series is indexed to 100 at…
- Founder as Agent Orchestrator
Claude Code / Cowork / Anthropic — the surfaces orchestration runs on
- The Future of Agent Interfaces
MCP's durable value is simple: it makes external systems agent-legible. A server exposes typed…
- Gemini Enterprise Agent Platform
The Google-side counterpart to Claude Code's and Codex's agent stacks — but where those entries…
- Harness Shrinkage as Models Improve
The harness — prompts, skills, scaffolding, mechanical verification — exists to compensate for what…
- HTML as the New Markdown
Claude Code — the product these HTML-artifact workflows run in, demonstrated from inside the Claude…
- Jarred Sumner
Everything Sumner reports here is case-study with an explicit disclosure: an Anthropic employee,…
- Latent Capability Overhang
Boris Cherny (YC interview, July 2026, practitioner-opinion) names the same gap from the product…
- Living Design System
Claude Code — extracts the design DNA from repos and renders design_system.html; Claude Design…
- Managers as ICs
Fiona Fung's "spicy" org change on Claude Code: every manager starts as an IC first and stays…
- MCP Tool Poisoning
Claude Code — named as a prominent MCP host; the class of agent this threat targets, and one of the…
- Entities — People, Orgs, Tools & Projects
Claude Code — Anthropic's agentic coding product; created by Boris Cherny late 2024;…
- Nate Parrott
Claude Code — was the sole product designer on Claude Code for VS Code; also the…
- Optimizer–Evaluator Decoupling
Claude Code v2.1.215 — the rule enforced by removing an affordance rather than by design. The…
- Orchestration vs Employee Framing: Reconciling the Founder's Playbook with HBR's Accountability Evidence
The playbook's lifecycle structure — Idea / MVP / Launch / Scale, each stage compressing what used…
- Parallel Agent Orchestration
Two numbers from it belong on this page. Fan-out width should vary with depth, and no fixed setting…
- Peter Steinberger
This is the seed of Loop Engineering — the shift from human-as-prompter to human-as-loop-designer.…
- Problem-Solution Fit Discipline
Idea-stage thesis: three defenses against premature building (time, resources, belief friction) all eroded; AI as devil…
- Recursive Self-Improvement
2025–2026 — Coding agents. Agents write and edit whole files on their own (Claude Code launches Feb…
- Same-Model Review Blindness
Two datasets of 500 pull requests each, one authored by Claude Code and one by Codex, identified by…
- Security Debt of Agent-Generated Code
Agent and language stratification (Figure 4, corpus average 38.9%): Copilot 45.5%, Claude Code…
- Standardize the Infrastructure, Not the Tools
The mechanism is an internal LLM proxy — a single gateway every AI request passes through before…
- Verification as the New Bottleneck
Fiona Fung's central claim from running Claude Code + Cowork engineering: for years, engineering…
- When Does Verification Quality Determine Whether AI Automation Works?
That is why Verification As The New Bottleneck is the org-level consequence of the Verifiability…
- Vibe Coding vs. Agentic Engineering
Vibe coding raises the floor (anyone builds); agentic engineering preserves the quality bar while going faster; ">10x a…
Related articles
- Harness Shrinkage as Models Improve
Prompt scaffolding shrinks each model release; Cat Wu's pruning discipline; Boris Cherny "100 lines of code a year from…
- Anthropic
AI safety company / vendor of Claude; mission-as-tiebreaker culture; ~30–40 PMs across teams; Mike Krieger leads Labs r…
- Open Questions Backlog
_456 actionable open questions across 205 pages · 107 predictions · 9 notes · 147 in progress · 69 watching (entities),…
- Boris Cherny
Creator of Claude Code at Anthropic; phone-driven workflow with hundreds of agents; primary advocate of `/loop` primiti…
- Verification as the New Bottleneck
Fiona Fung: coding is no longer the bottleneck — verification, review, maintenance are; shift-left; TDD loses its tax;…
