这是一篇由原始材料转换而来的阅读页,保留了源文件的主要结构,并补充了可追溯的来源说明与链接。
摘要
Use this when the repository's main bottleneck is stale or fragmented knowledge.
openaimarkdownarticle
Framework 4 — Documentation Garden / Knowledge Reliability
Use this when the repository's main bottleneck is stale or fragmented knowledge.
Intent
Keep repo knowledge navigable, fresh, and mechanically checkable so agents do not optimize against dead documentation.
Core components
- short
AGENTS.md - structured
docs/tree - cross-link conventions
- freshness / owner / status metadata where helpful
- periodic doc-gardening runs
- CI checks for broken references and missing entry points
Recommended folders
docs/index.mddocs/architecture/docs/runbooks/docs/debt/plans/done/for historical decisions
Typical loop
- Find stale or unlinked docs
- Compare docs against code paths and scripts
- Repair or archive obsolete docs
- Improve discoverability from docs/index.md
- Encode repeat problems into checks or templates
Best fit
- mature repos
- multi-team systems
- agent-heavy environments where hidden knowledge is the main blocker
来源与参考
源文件: openai/framework-doc-garden/README.md
来源目录: /srv/project/harness-engineering
继续阅读
Framework 3 — Bug Reproduction / Fix / Regression FactoryUse this when the main workload is diagnosing and fixing bugs rapidly.Framework 2 — Long-Running Multi-Session EngineeringUse this when a task will span many agent sessions or many context windows.OpenAI Harness Engineering — 读后感 / 总结 / 可迁移经验也就是: 给什么文档 如何组织知识 如何定义边界 如何给验证闭环 如何给清理与治理机制