docs: expand .cards — vision, comparisons, business rationale, open questions
CI / check (pull_request) Successful in 2m21s

26 cards covering:
- Project philosophy (session isolation, process discipline, dissipative structure)
- Comparisons (vs skill, vs dynamic workflow)
- Business rationale (FTE vs vendor, OPC, switching cost)
- Learning model (memory + skill + workflow)
- Self-improvement (reflective workflow, eval)
- Open questions (workflow granularity, human-in-the-loop)
This commit is contained in:
2026-06-07 11:43:36 +00:00
parent f697aec3e7
commit 04e2b5b8a7
21 changed files with 485 additions and 0 deletions
+22
View File
@@ -0,0 +1,22 @@
---
title: "Three Learning Carriers — Memory, Skill, Workflow"
created: "2026-06-07"
source: "openclaw-xiaomo"
tags: [architecture, pattern]
category: "architecture"
links:
- agent-as-graduate
- skill-vs-workflow-different-layers
- when-skill-is-not-enough
- opc-why-fte-agents-matter-most
---
完整的 FTE agent 需要三层能力载体,缺一不可:
- **记忆强化知识** — 事实性积累。"这个客户偏好什么"、"上次这个 API 出过什么问题"。让 agent 越来越懂你的业务上下文。
- **Skill 强化技能** — 单个 session 内的方法论。"怎么做 code review"、"怎么写测试"。提升每个环节的执行质量。
- **Workflow 强化纪律** — session 之间的协作结构。"做一个 feature 必须经过哪些步骤"。保证流程可靠性和多视角制衡。
三者互补:光有 workflow 没有 skill,每个环节执行质量差;光有 skill 没有 workflow,能力强但不守规矩;光有记忆没有前两者,知道很多但不会做事。
出厂能力同样重要——底座模型 + 技能包决定毕业生的起点。名校博士和高中毕业,出厂就不同。但出厂能力是必要不充分的,FTE 的长期价值靠三层载体的持续积累。