<!-- registry: A8, A9 -->

参考来源

本课程的关键论断都以下列来源为依据。正文用 [^Sn] 引用，具体引文摘录在每条下方。

S1 — Effective context engineering for AI agents — Anthropic Engineering

URL: https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents

• authority: authoritative-guide

Anthropic 工程团队关于上下文工程的权威长文，是本课程的核心依据：定义上下文工程与 Prompt 工程的关系、注意力预算与上下文腐化、系统提示的「高度」、工具与示例作为上下文、即时检索与渐进式披露、压缩、结构化笔记、子代理架构这几乎全部主题都出自此文。

关键引用：

"we view context engineering as the natural progression of prompt engineering" "Prompt engineering refers to methods for writing and organizing LLM instructions for optimal outcomes" "the set of strategies for curating and maintaining the optimal set of tokens (information) during LLM inference" "An agent running in a loop generates more and more data that could be relevant for the next turn of inference" "LLMs have an "attention budget" that they draw on when parsing large volumes of context" "Every new token introduced depletes this budget by some amount" "as the number of tokens in the context window increases, the model's ability to accurately recall information from that context decreases" "some models exhibit more gentle degradation than others, this characteristic emerges across all models" "These factors create a performance gradient rather than a hard cliff" "context, therefore, must be treated as a finite resource with diminishing marginal returns" "agents that operate over multiple turns of inference and longer time horizons" "engineers hardcoding complex, brittle logic in their prompts to elicit exact agentic behavior" "vague, high-level guidance that fails to give the LLM concrete signals for desired outputs" "specific enough to guide behavior effectively, yet flexible enough to provide the model with strong heuristics" "tools should be self-contained, robust to error, and extremely clear with respect to their intended use" "building tools that are well understood by LLMs and have minimal overlap in functionality" "returning information that is token efficient and by encouraging efficient agent behaviors" "stuff a laundry list of edge cases into a prompt" "curate a set of diverse, canonical examples that effectively portray the expected behavior of the agent" "maintain lightweight identifiers (file paths, stored queries, web links, etc.)" "the metadata of these references provides a mechanism to efficiently refine behavior" "allows agents to incrementally discover relevant context through exploration" "retrieving some data up front for speed, and pursuing further autonomous exploration at its discretion" "CLAUDE.md files are naively dropped into context up front, while primitives like glob and grep" "taking a conversation nearing the context window limit, summarizing its contents, and reinitiating a new context window" "passing the message history to the model to summarize and compress the most critical details" "preserves architectural decisions, unresolved bugs, and implementation details while discarding redundant tool outputs" "the agent regularly writes notes persisted to memory outside of the context window" "Like Claude Code creating a to-do list, or your custom agent maintaining a NOTES.md file" "Claude playing Pokémon demonstrates how memory transforms agent capabilities in non-coding domains" "specialized sub-agents can handle focused tasks with clean context windows" "returns only a condensed, distilled summary of its work (often 1,000-2,000 tokens)" "the detailed search context remains isolated within sub-agents" "compaction, structured note-taking, and multi-agent architectures" "maintain coherence, context, and goal-directed behavior over sequences of actions"

S2 — Building Effective AI Agents — Anthropic Engineering

URL: https://www.anthropic.com/engineering/building-effective-agents

• authority: authoritative-guide

Anthropic 关于 Agent 模式的权威文章。本课用它支撑 Agent 的循环定义、自主性的成本与误差累积、以及「只在复杂度确实改善结果时才增加」这条分寸原则。

关键引用：

"They are typically just LLMs using tools based on environmental feedback in a loop." "The autonomous nature of agents means higher costs, and the potential for compounding errors." "The LLM will potentially operate for many turns, and you must have some level of trust in its decision-making." "you should consider adding complexity only when it demonstrably improves outcomes."

S3 — How we built our multi-agent research system — Anthropic Engineering

URL: https://www.anthropic.com/engineering/multi-agent-research-system

• authority: authoritative-guide

Anthropic 多 Agent 研究系统的工程复盘，提供子代理上下文隔离与 token 经济学的第一手数据：并行子代理各占独立上下文窗口、子代理充当「智能过滤器」为主代理压缩发现、以及 Agent 与多 Agent 系统的 token 用量倍数。

关键引用：

"Subagents facilitate compression by operating in parallel with their own context windows" "distributes work across agents with separate context windows to add more capacity for parallel reasoning" "agents typically use about 4× more tokens than chat interactions" "multi-agent systems use about 15× more tokens than chats" "Multi-agent systems work mainly because they help spend enough tokens to solve the problem." "token usage by itself explains 80% of the variance, with the number of tool calls and the model choice" "condensing the most important tokens for the lead research agent" "The essence of search is compression: distilling insights from a vast corpus." "the subagents act as intelligent filters" "agents summarize completed work phases and store essential information in external memory" "agents can spawn fresh subagents with clean contexts while maintaining continuity through careful handoffs"

S4 — Best practices for Claude Code — Claude Code Docs

URL: https://code.claude.com/docs/en/best-practices

• authority: official-docs

Claude Code 官方最佳实践页（anthropic.com/engineering/claude-code-best-practices 已 308 重定向至此）。提供上下文管理在真实产品里的落地形态：CLAUDE.md 的取舍纪律、/clear 与自动压缩、子代理调查隔离主上下文、以及「上下文是最重要的资源」这条总纲。

关键引用：

"Claude's context window fills up fast, and performance degrades as it fills." "The context window is the most important resource to manage." "This matters since LLM performance degrades as context fills." "CLAUDE.md is a special file that Claude reads at the start of every conversation." "CLAUDE.md is loaded every session, so only include things that apply broadly." "Claude loads them on demand without bloating every conversation." "Keep it concise. For each line, ask: "Would removing this cause Claude to make mistakes?" If not, cut it." "Bloated CLAUDE.md files cause Claude to ignore your actual instructions!" "If your CLAUDE.md is too long, Claude ignores half of it because important rules get lost in the noise." "reset context between unrelated tasks. Long sessions with irrelevant context can reduce performance." "Run /clear between unrelated tasks to reset the context window entirely" "During long sessions, Claude's context window can fill with irrelevant conversation, file contents, and commands." "Claude Code automatically compacts conversation history when you approach context limits, which preserves important code and decisions while freeing space." "Since context is your fundamental constraint, subagents are one of the most powerful tools available." "Subagents run in separate context windows and report back summaries" "Scope investigations narrowly or use subagents so the exploration doesn't consume your main context." "A clean session with a better prompt almost always outperforms a long session with accumulated corrections."

S5 — The 2026 Agent Engineering Roadmap — GitHub (codejunkie99/agent-roadmap-2026)

URL: https://github.com/codejunkie99/agent-roadmap-2026

• authority: blog

以「harness 工程」为主线的社区开源路线图（AI 辅助撰写）。本课只取其框架性论点：上下文管理是 harness 的组件之一、以及它对上下文工程的一句话定义。注意：其中的百分比阈值、token 数字、基准分数、薪资等具体数字一律不作为事实引用；「Prompt engineering is dead as a standalone skill in 2026」是该路线图的观点性断言，引用时必须注明出处立场，课程正文优先采用 S1 更审慎的「自然演进」表述。

关键引用：

"Same model, different harness, completely different result." "the harness is the union of:" "context engineering: deciding what tokens are in front of the model at every step of the loop" "Prompt engineering is dead as a standalone skill in 2026."