Skip to main content
reopt Handbook
reopt Handbook
Harness Engineering

Problem and Structure

Foundations of Harness EngineeringRepo-Readable SystemsThe Five Elements of a HarnessEngineering Mechanics

Evaluation and Interpretation

Evaluation Loop DesignExternal Case Comparison

Case Studies

Case: OpenAICase: AnthropicCase: TossCase: gstackCase: revfactory/harness

Domain Playbooks

Domain PlaybooksScenario: Frontend TeamScenario: Platform TeamScenario: Payments and Settlement TeamScenario: AI Product TeamWhy You Eventually Need Your Own Harness

Rollout and Operations

Team Harness Rollout StrategyTeam Harness Design ChecklistOperations: Entropy and Garbage Collection

Appendix

Verification ReportUpdates
Handbook›Harness Engineering›Verification Report
한국어English

Verification Report

Link, consistency, source, and static validation report for the Harness Engineering handbook.

This document records structure, source, cross-link, and handbook app validation for the English Harness Engineering handbook.

Verification Baseline

2026-05-23

Scope

ItemStandard
Document structuremeta.json matches actual MDX files
Content consistencyCore claims and chapter flow do not conflict
External evidenceOpenAI, Anthropic, Toss, gstack, and revfactory/harness claims are checked
Cross-linksRelated handbook links are valid
App validationhandbook registry, typecheck, and build pass

Method

  1. Compared apps/handbook/content/books/en/harness-engineering/meta.json with the MDX file list.
  2. Compared chapter claims against the source material.
  3. Checked related links to LLMOps/AgentOps, Codex, Claude Code, orchestration, and documentation books.
  4. Used OpenAI developer docs MCP and official OpenAI sources for OpenAI items.
  5. Cross-checked Anthropic engineering/news sources and Claude Code / Managed Agents official docs.
  6. Checked gstack and revfactory/harness README state and GitHub metadata during the Korean baseline update.
  7. Ran static validation.

Result Summary

ItemResult
meta.json and MDX files23 pages aligned
Structure flowPass
External evidence connectionPass
Cross-linksPass
pnpm --filter handbook run check:books-registryPass
pnpm --filter handbook run typecheckPass
pnpm --filter handbook run buildPass

Core Sources

SourceDateUse in this book
OpenAI, Harness Engineering2026-02-11agent-readable repo, short AGENTS.md, structured docs, observability, garbage collection
OpenAI, The next evolution of the Agents SDK2026-04-15model-native harness, native sandbox execution, MCP/skills/AGENTS.md/shell/apply_patch primitives
OpenAI API Changelog2026-05-06 / 2026-05-19TypeScript sandbox agents, open-source harness, Secure MCP Tunnel
OpenAI Developers plugin for Codex2026-05-07OpenAI Platform access, API key setup, troubleshooting as plugin surface
OpenAI, Work with Codex from anywhere2026-05-14mobile/remote connection, approvals, hooks, enterprise environment
OpenAI Agents SDK / Sandbox / Codex docsRead baseline 2026-05-23sandbox capability, hooks lifecycle, remote connections
Anthropic, Harness design for long-running application development2026-03-24planner/generator/evaluator and load-bearing scaffolding
Anthropic, Claude Code auto mode2026-03-25prompt-injection probe, transcript classifier, trust boundary, denial fallback
Claude Code permission / auto mode docsRead baseline 2026-05-23permission modes, protected paths, classifier order, trusted infrastructure
Anthropic, Scaling Managed Agents2026-04-08session/harness/sandbox split, durable event log, credential vault, MCP proxy
Claude Managed Agents overview / MCP connector docsRead baseline 2026-05-23agents, environments, sessions, events, MCP auth/vault separation
Anthropic, Agents for financial services2026-05-05domain templates, skills/connectors/subagents, per-tool permissions, audit log
Toss harness article2026-02-26frictionless harness, executable SSOT, domain layer, HITL
gstack READMERead baseline 2026-05-23specialists, power tools, agent hosts, team mode, QA, checkpoint, learning
revfactory/harness READMERead baseline 2026-05-23L3 Meta-Factory, Team-Architecture Factory, architecture patterns, A/B caveat

Synthesized Claims

ClaimEvidence basis
Harnesses are work-system design, not prompt tricksOpenAI repo/observability + Anthropic evaluation + Toss system rollout
Generic harnesses are starting pointsToss domain layers + gstack make-it-yours posture + revfactory domain teams
Operations and cleanup are part of the harnessOpenAI entropy and doc gardening view
Harness primitives are being productized, but domain design remainsOpenAI primitives + gstack/revfactory domain workflows
Auto approval is a policy layer, not a human-review replacementAnthropic auto mode classifier/trust-boundary structure
Long-running agent runtime should use durable session logsAnthropic Managed Agents session/harness/sandbox split

Limitations

Scope

This report reflects the 2026-05-23 source baseline. External tools and docs can change quickly; interpretation changes should be recorded in updates.mdx.

Related docs

External Case Comparison

Compare OpenAI, Anthropic, Toss, gstack, and revfactory/harness by input, state, verification, and rollout.

Updates

Change log for the Harness Engineering handbook

Verification Report

LLMOps and AgentOps in Production · Structure, link, metric, and logic verification for LLMOps and AgentOps in Production

Verification Report

Agentic Documentation · Source, link, consistency, and build verification for the Agentic Documentation handbook.

Updates

LLMOps and AgentOps in Production · Changelog for LLMOps and AgentOps in Production

Operations: Entropy and Garbage Collection

Explain how harnesses decay and how teams keep docs, workflows, permissions, hooks, and runtime surfaces current.

Updates

Change log for the Harness Engineering handbook

On this page

ScopeMethodResult SummaryCore SourcesSynthesized ClaimsLimitations