Agent Harness Survey — 학습 리소스 인덱스
Agent Harness Engineering: A Survey 를 본격 공부할 때 함께 보면 좋은 1차/2차 자료 모음.
1. 1차 자료 (Primary Sources)
| 종류 | 링크 | 비고 |
|---|---|---|
| Project page | https://picrew.github.io/LLM-Harness/ | 단일 페이지 요약 + 핵심 다이어그램 |
| PDF (project) | https://picrew.github.io/LLM-Harness/main.pdf | 본문 PDF |
| PDF (OpenReview) | https://openreview.net/pdf?id=eONq7FdiHa | 동일 PDF, 인용 가능 |
| OpenReview forum | https://openreview.net/forum?id=eONq7FdiHa | 토론·메타데이터 |
| GitHub (paper repo) | https://github.com/Picrew/LLM-Harness | docs/ + README, MIT/CC BY-SA 4.0 |
| Awesome list | https://github.com/Picrew/awesome-agent-harness | 207 entries, 9 categories, 87.9% GitHub-backed (2026-05-25 verified) |
| HF dataset | https://huggingface.co/datasets/ChenLiu1996/Agent-Harness-Engineering | 코딩된 프로젝트 데이터셋 |
| BibTeX | li2026agentharness | OpenReview PDF URL 인용 |
2. Awesome-Agent-Harness 카테고리 인덱스
207개 항목을 ETCLOVG에 맞춰 다시 보면:
| Awesome 카테고리 | Entries | ETCLOVG 매핑 |
|---|---|---|
| Harness Architecture & Orchestration | 27 | L |
| Context & Working-State Engineering | 10 | C |
| Execution Substrates & Sandboxing | 23 | E |
| Protocols, Tool Interfaces & Agent Contracts | 14 | T |
| Evaluation Harnesses & Benchmarks | 24 | V |
| Observability & Reliability Operations | 14 | O |
| Guardrails, Security & Governance | 16 | G |
| Reference Harness Implementations | 50 | (혼합) |
| Essential Readings & Ecosystem Maps | 29 | 메타 |
카테고리는 awesome list 분류 / 서베이 본문 통계는 “primary layer” 카운트 기준(서로 정의가 약간 다름).
3. Featured Harness Blogs (저자들이 핵심으로 꼽은 글)
프로덕션 회사의 1차 글:
- Anthropic — Scaling Managed Agents (“brain ↔ hands decoupling”) — meta-harness architecture for long-horizon agents
- Anthropic — Claude Code auto mode — classifier-backed approval delegation
- OpenAI — Harness engineering field report — building reliable agent-first software via harness constraints + verification
- Anthropic — Building Effective AI Agents — workflows vs autonomous agents
- Anthropic — Writing effective tools for AI agents — tool interface design
- Anthropic — Effective harnesses for long-running agents — state / resumability / reliability
- Anthropic — Harness design for long-running application development — follow-up
- LangChain — Improving Deep Agents with harness engineering — 하네스만 바꿔도 +13.7점 증거
- LangChain — Evaluating Deep Agents: Our Learnings — stateful 장기 에이전트 평가
- Inngest — Your Agent Needs a Harness, Not a Framework — reliability-first 인프라
→ 우리 vault 미러: 2026-04-30-improving-deep-agents-with-harness-engineering, harness
4. 본문에서 인용된 시대별 시스템
| Era | Systems |
|---|---|
| 2022–2023 (ReAct) | ReAct, AutoGPT, BabyAGI |
| 2023–2024 (tool & multi-agent) | Gorilla, ToolLLM, Toolformer · CAMEL, ChatDev, MetaGPT, Mixture-of-Agents |
| 2023–2024 (benchmarks) | SWE-bench, AgentBench, WebArena, GAIA |
| 2024–2025 (protocols) | MCP, A2A |
| 2025–2026 (harness era) | LangChain Deep Agents, Anthropic Managed Agents, OpenAI harness engineering, Picrew survey |
5. ETCLOVG 7계층 sub-page (이 vault)
- etclovg-e-execution
- etclovg-t-tooling
- etclovg-c-context
- etclovg-l-lifecycle
- etclovg-o-observability
- etclovg-v-verification
- etclovg-g-governance
6. NotebookLM 노트북 ✅ 생성 완료
- 노트북: Agent Harness Engineering — A Survey (2026)
- Notebook ID:
8e64ee73-5358-4ec2-a8d1-3b8d322c5228 - PDF 사본 (vault):
assets/papers/2026-li-agent-harness-engineering-survey.pdf(3.3MB)
추가된 소스 (4개):
| Source | Type | ID |
|---|---|---|
2026-li-agent-harness-engineering-survey.pdf | file | 8bae14d0-be73-463c-b5e1-20f61d76bb78 |
| Project page (picrew.github.io/LLM-Harness) | url | 7f808eff-f067-473d-8213-8760404285c8 |
| Awesome list (Picrew/awesome-agent-harness) | url | 4b27d8ee-3a8e-426e-8d20-64ac88d16072 |
| OpenReview forum | url | 0c6e2baa-003e-4759-bde2-f231d65e9af3 |
처음 던지면 좋은 질문 (한국어):
- “ETCLOVG와 기존 6-component framework들의 정확한 차이를 표로 정리해줘. 어떤 framework들이 비교 대상인가?”
- “harness coupling problem을 보여주는 구체 사례를 본문에서 직접 인용으로 가져와줘.”
- “Open Problem 2 (reliable state in long-running agents)가 권장하는 durable artifacts 패턴의 실제 구현 사례가 본문에 있나?”
- “trace-native failure diagnosis가 가장 잘 구현된 오픈소스 프로젝트가 어떤 것이라고 본문에서 평가하나?”
- “Anthropic의 ‘managed agents’ 글이 ETCLOVG 중 어느 레이어들과 가장 강하게 연결되는가?”
- “이 서베이가 ‘agent framework → agent platform’ 이동을 말하는 부분을 그대로 인용해줘.”
자동화가 다시 가능해지면:
nlm login # 재인증
# 그 후 Claude에게: "NotebookLM에 노트북 만들고 assets/papers/2026-li-...pdf 업로드해줘"7. 함께 읽을 우리 vault 노트
- harness — entity 페이지, 8 key concepts
- moc-ai-agents-harness — MOC
- 2026-04-19-harness-is-everything
- 2026-04-19-anatomy-of-agent-harness
- 2026-04-30-improving-deep-agents-with-harness-engineering
- 2026-04-16-agent-skills-realistic-benchmark-gap
- 2026-05-21-agent-harness-landscape
- 2026-05-25-threads-geun-daeng-harness-benchmark