好文集ChatGPT Plus¥119

Context vs Prompt vs Loop vs Harness Engineering: The Four-Layer Agent Stack

文章目录

部分视频或图片暂未收录,已在正文保留提示;需要时可查看原文。查看原文 ↗

Prompt, context, loop, and harness engineering are four different layers. This guide maps the stack, shows where each lever lives, and links to deep dives on Claude Code, MCP, and production agent design


Context engineering, Loop engineering, Harness engineering = upgrades to "Prompt engineering"

文章配图

They are not interchangeable. They are four layers of the same stack, each with different units of work, failure modes, and tools

  • Prompt engineering asks: How do I word this message?

  • Context engineering asks: What does the model see on this call?

  • Loop engineering asks: What autonomous workflow repeats until a goal is met?

  • Harness engineering asks: What code runs the loop, tools, and verification reliably?

Confusing them is expensive. A team that rewrites prompts when the harness has no verification step will never fix silent failure loops. A team that builds a sophisticated harness with vague goals will burn tokens forever

The same model can also consume radically different token totals across harnesses, but token count alone is not a quality score

文章配图

This guide maps the full stack - with diagrams, diagnostics, and detailed visuals


01. TL;DR - the four layers at a glance

LayerUnit of workYou design…Typical artifactsWhen it dominates
PromptOne messageWording, format, CoT, few-shotSystem prompt text, user templateSingle-turn tasks, prototyping
ContextOne model callFull context packageCLAUDE.md, RAG chunks, tool list, history prune rulesMulti-turn agents, RAG, cost control
LoopEntire runTrigger → goal → verify → memory/goal, cron, Agent View, triage specsAutonomous coding, scheduled agents
HarnessRuntime executionTool sandbox, retries, checkpointsClaude Code, LangGraph, custom orchestratorProduction reliability, benchmarks

Nesting rule: Harness implements loops → each loop step assembles context → context contains prompts


02. The stack diagram

┌─────────────────────────────────────────────────────────────┐
│  HARNESS ENGINEERING                                        │
│  Runtime: tool exec, sandbox, retries, checkpoints, logs    │
│  ┌───────────────────────────────────────────────────────┐  │
│  │  LOOP ENGINEERING                                     │  │
│  │  Workflow: trigger, goal, actions, verification, memory │  │
│  │  ┌─────────────────────────────────────────────────┐  │  │
│  │  │  CONTEXT ENGINEERING (per iteration)            │  │  │
│  │  │  Assembly: history, RAG, tools, CLAUDE.md, MCP  │  │  │
│  │  │  ┌───────────────────────────────────────────┐  │  │  │
│  │  │  │  PROMPT ENGINEERING (messages inside)     │  │  │  │
│  │  │  │  Wording: role, format, constraints, CoT  │  │  │  │
│  │  │  └───────────────────────────────────────────┘  │  │  │
│  │  └─────────────────────────────────────────────────┘  │  │
│  └───────────────────────────────────────────────────────┘  │
└─────────────────────────────────────────────────────────────┘
         ▲                              ▲
         │                              │
    Boris Cherny:                  Andrej Karpathy:
    "build loops"                  "context engineering"
    (loop + harness)               (context + prompt)

Boris Cherny about self-verification loops:

嵌入推文(联网查看)

Boris Cherny: "Two of the most powerful features in Claude Code: /loop and /schedule":

嵌入推文(联网查看)

Andrej Karpathy about "context engineering":

嵌入推文(联网查看)

Think of it like web development:

Agent stackWeb analogy
PromptCopy on a single button label
ContextFull page layout + data fetched for this view
LoopUser journey / multi-step checkout flow
HarnessBrowser, server, DB, auth, error handling

03. Layer 1 - Prompt engineering (innermost)

Boris Cherny: "I am surprised that people are surprised this is how I prompt Claude."

嵌入推文(联网查看)

Definition: Optimizing the text of individual messages, usually the system prompt and current user turn, to elicit better model behavior

Techniques: Chain-of-thought, few-shot examples, role assignment, JSON schema instructions, temperature and sampling controls, system prompt structure

Unit of work: One message pair (system + user) or one turn in a chat

文章配图

Example - prompt-level fix:

Before (vague):
Summarize this doc.

After (prompt-engineered):
You are a staff engineer writing release notes for developers.
Output: 3 bullets, each ≤25 words, past tense, no marketing language.
If the doc lacks version numbers, say "Version unclear" — do not invent one.

That improves a single call when the right document is already in context.

When prompt engineering is enough:

  • One-shot translation, classification, or formatting

  • Early prototyping before you know the workflow shape

  • Failures are clearly about misunderstood instructions, not missing files

When it is not enough:

  • The model "doesn't know" your codebase → context problem

  • The task needs 40 tool calls → loop problem

  • Tool calls hang or duplicate writes → harness problem

context vs prompt engineering:

文章配图


04. Layer 2 - Context engineering (per call)

Definition: Designing the full package the model conditions on for each API call

In mid-2026, Karpathy popularized the idea that the real challenge lies in curating the context window:

嵌入推文(联网查看)

What lives in the context package

ComponentContext decisionNot a prompt decision
System promptYes (wording)Also placement — constraints at top
Conversation history—Keep, summarize, or drop turns
Retrieved docs—Which chunks, how many tokens
Tool definitions—Expose 3 tools or 30?
Tool outputs—Full stdout vs truncated summary
CLAUDE.md—Always-loaded project rules
SKILL.md—Load on trigger only
MCP results—Live data injected at query time

The four context levers

  1. Content selection = does this token earn its place?

  2. Structure and ordering = constraints first; docs before user message

  3. Token budget = protect budget for variable content

  4. Cache placement = stable prefix (system + tools) before cache breakpoint

Example - context-level fix (same user message)

[SYSTEM — always loaded]
Project: [........]. Package manager: pnpm. Tests: pnpm test --filter web.
Never edit apps/mobile without explicit ask.

[RETRIEVED — grep hit, apps/web/lib/pathway-data.ts, lines 40-95]
{relevant_snippet_only}

[TOOLS — this task only]
Read, Edit, Bash(pnpm test:*)

[USER]
Fix the failing pathway progress test.

The user message is boring on purpose. Quality comes from what surrounds it

Context engineering surfaces in Claude Code:

SurfaceLayerLoads when
~/.claude/CLAUDE.mdContextEvery session
./CLAUDE.mdContextProject session
SKILL.mdContextTask trigger
MCP serversContext + HarnessTool call time
Context Mode / sandbox MCPContext + HarnessIsolated file reads

Full stack breakdown:

文章配图

When context engineering dominates

  • RAG pipelines and codebase Q&A

  • Sessions beyond ~10 turns (history pollution)

  • Tool-selection errors (too many tools exposed)

  • Cost/latency (200k context that should be 40k)


05. Layer 3 - Loop engineering (workflow)

Definition: Designing the autonomous workflow that decides when to call the model, what goal must be satisfied, and how to know the run is done without you typing each turn

Addy Osmani popularized loop engineering in June 2026, building on Boris Cherny at Anthropic:

"I don't prompt Claude anymore. I have loops that are running. They're the ones that are prompting Claude and figuring out what to do. My job is to write loops"

嵌入推文(联网查看)

Addy Osmani's guide:

嵌入推文(联网查看)

The loop still contains prompts = something generates them each iteration. You stop being that something

The five loop components

ComponentQuestion it answersBad design symptom
TriggerWhat starts the run?You still paste prompts manually
GoalWhat verifiable state ends it?Agent "finishes" but tests fail
ActionsWhat tools can it use?Agent can't reach GitHub/DB
VerificationHow do we check progress?Infinite loops or premature stop
MemoryWhat persists across steps?Re-reads same files, repeats edits

Example - loop spec (not a prompt)

name: morning_p1_triage
trigger: cron "0 8 * * 1-5"
goal: zero open GitHub issues labeled P1 without assignee
actions: [github_mcp.list_issues, github_mcp.comment, github_mcp.assign]
verify: script checks assignee field on all P1 issues
memory: log file of triaged issue IDs this week
max_iterations: 20
human_gate: none  # read/write on issues only

No clever phrasing - workflow architecture

Loop engineering vs prompt engineering:

DimensionPromptLoop
Who drives turnsYouSystem
DurationSecondsMinutes to hours
OutputTextVerified outcome
Leverage1×10–100×
Primary skillPhrasingSystems design

When loops fail (and it's not the prompt)

SymptomLoop fix
Runs foreverAdd max_iterations + no-progress detector
Stops after one fileTighten goal; add test verification
Does wrong work confidentlyGoal too vague — use verifiable criteria
Repeats same editMemory / checkpoint missing

06. Layer 4 - Harness engineering

Definition: Building or configuring the orchestration code that executes loops: parsing model output, calling tools safely, managing retries, assembling context each turn, and enforcing exit conditions

Boris Cherny and Anthropic engineers use harness engineering for the systems that prompt Claude iteratively: observe, plan, act, reflect => over hours

Addy Osmani's guide:

嵌入推文(联网查看)

The agent harness is the concrete artifact:

Goal in → context assembly → model call → parse → execute tools
       → capture results → verify → loop or exit → result out

Harness components

ComponentWhat it doesLoop vs harness
Task definitionConverts goal to first promptLoop designs; harness encodes
Context managerPrunes history, injects memoryContext rules; harness implements
Tool executorSandboxed bash, MCP, file I/OHarness
Loop controllerIteration limits, exit signalsHarness
VerificationRuns tests, scripts, diff checksLoop specifies; harness runs
Retry / checkpointIdempotent retries, resumeHarness
ObservabilityLogs, traces, cost metersHarness

Why harness beats model upgrades on benchmarks

LangChain's Deep Agents team reported gains on Terminal-Bench 2.0 from harness changes alone → same underlying model. The pattern generalizes:

Better harness on the same model > same harness on a better model - for many agentic tasks

Reason: the model only sees what the harness assembles and only gets the retries the harness allows

Products as harnesses

ProductHarness features
Claude CodeTools, hooks, permission modes, sessions, subagents
Cursor / CodexIDE integration, model routing, agent modes
LangGraphStateful graphs, checkpoints, human-in-the-loop
OpenCodeOpen-source coding agent harness
CustomYour retry logic, your verification scripts

07. How the layers interact on one real task

Task: "Migrate our auth module to Clerk and make all tests pass."

Prompt layer:

Migrate auth to Clerk. Make tests pass.

Works for a toy repo. Fails on a monorepo in one shot

+ Context layer

  • Load CLAUDE.md with package manager, test command, auth file map

  • Retrieve only apps/web/lib/auth/* via grep - not whole repo

  • Expose Edit + Bash(test) tools only

Each iteration of the agent sees a sane package

+ Loop layer

trigger: developer runs /goal
goal: pnpm test --filter web exits 0 AND auth routes use Clerk SDK
verify: test command + grep for legacy auth imports
max_iterations: 50
memory: PROGRESS.md updated each checkpoint

Agent runs until verified → not until it says "done"

+ Harness layer

  • Sandbox bash; block rm -rf

  • Hooks run lint after every edit

  • Checkpoint git stash every 10 turns

  • Retry transient API failures

  • Log token usage per step

  • Human gate before editing production env files

Production shipping requires all four


08. Diagnostic - which layer is broken?

Use this when an agent underperforms:

SymptomLikely layerFirst fix
Model misunderstands instruction wordingPromptRewrite system prompt; add few-shot
Model lacks facts not in prompt textContextAdd RAG, CLAUDE.md, or file read
Model ignores constraints mid-sessionContextMove constraints to top; repeat before user msg
Wrong tool selectedContextReduce tool surface; improve schemas
Quality degrades after turn 15ContextSummarize/prune history
Never completes taskLoopAdd verifiable goal + test verification
Completes but wronglyLoopStrengthen verify step
Repeats same actionLoopAdd memory + no-progress detector
Duplicate emails / double writesHarnessIdempotent retries, checkpoints
Hangs on subprocessHarnessTimeouts, kill switches
Can't debug what happenedHarnessStructured logging, traces

Fix bottom-up within a layer, outer layers first across layers: if verification never runs, no prompt edit helps


09. The 2026 career map

Role focusPrimary layersSecondary
Content / marketing AIPrompt, Context—
Support bot with KBContext, PromptLoop (escalation)
Internal coding assistantContext, LoopHarness (CI integration)
Autonomous coding agentLoop, HarnessContext
Platform / agent infraHarnessLoop, Context

Prompt engineering is table stakes → like knowing SQL if you build backends

Context engineering is required for any agent that reads your data

Loop engineering is required when you want autonomy measured in hours, not seconds

Harness engineering is required when failure has a cost → money, data, reputation


10. Practical learning path

文章配图


Bottom line

Prompt, context, loop, and harness engineering are not four names for the same job

They are four layers of one stack:

  • Prompt = message wording

  • Context = per-call assembly

  • Loop = autonomous workflow

  • Harness = reliable execution

Karpathy named the context crisis. Cherny named the loop shift. Production teams name the harness when benchmarks move without new models

When someone says "we need better prompts" on a long-running agent, ask: "Which layer is actually failing?"

The answer determines whether you edit a paragraph, redesign retrieval, rewrite the goal spec, or fix retry logic

Get the layer right → then optimize inward


Save this so you don't lose it

***Follow @beamnxw for more alpha ***

*=> *my substack

*=> *my telegram channel

=> my youtube

=> my github

=> my github works

文章配图

全文完
查看原文