← Back to English

vibecosystem

基于 Claude Code 构建的 AI 软件团队。

138 个 Agent。296 项 Skill。74 个 Hook。零人工干预。

概览

指标	数量
Agents	138
Skills	296
Hooks	74
Rules	20
人工操作	0

这是什么？

vibecosystem 将 Claude Code 变成了一支完整的 AI 软件团队。不是一个助手，而是由 138 个专业 agent 组成的团队 -- 它们负责规划、构建、审查、测试，并从每一个错误中学习。

无需自定义模型。无需自定义 API。仅凭 Claude Code 的 hook + agent + rules 系统，发挥到极致。

全景图

核心功能

1. 自学习流水线

每一个错误都会自动变成规则。

Error happens → passive-learner captures pattern
→ consolidator groups & counts
→ confidence >= 5 → auto-inject into context
→ 10x repeat → permanent .md rule file

无需人工介入。系统自己编写自己的规则。

2. Agent Swarm

只需说"添加一个新功能"，20 多个 agent 就会在 5 个阶段中协同启动。

Phase 1 (Discovery):    scout + architect + project-manager
Phase 2 (Development):  backend-dev + frontend-dev + devops + specialists
Phase 3 (Review):       code-reviewer + security-reviewer + qa-engineer
Phase 4 (QA Loop):      verifier + tdd-guide (max 3 retry → escalate)
Phase 5 (Final):        self-learner + technical-writer

3. 自适应 Hook 加载

74 个 hook 并非同时运行。用户意图决定哪些 hook 被触发。

"fix the bug"      → compiler-in-loop + error-broadcast     ~2,400 tok
"add api endpoint"  → edit-context + signature-helper + arch  ~3,100 tok
"explain this code" → (nothing)                               ~800 tok

4. Dev-QA Loop

每个任务都必须通过质量关卡：

Developer implements → code-reviewer + verifier check
→ PASS → next task
→ FAIL → feedback to developer, retry (max 3)
→ 3x FAIL → escalate (reassign / decompose / defer)

5. Canavar Cross-Training

当一个 agent 犯错时，整个团队都会从中学习。

Agent error → error-ledger.jsonl → skill-matrix.json
→ All agents get the lesson at session start
→ Team-wide error prevention

6. Assignment Matrix

74 行的路由表：每种任务类型都映射到最合适的 agent。

GraphQL API      → graphql-expert  (backup: backend-dev)
Kubernetes       → kubernetes-expert (backup: devops)
DDD modeling     → ddd-expert      (backup: architect)
Bug reproduction → replay          (backup: sleuth)
... 70 more rows

架构

┌─────────────────────────────────────────────────────────┐
│                    Claude Code                          │
│                                                         │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐              │
│  │  Hooks   │  │  Agents  │  │  Skills  │              │
│  │  (74)    │→ │  (138)   │← │  (296)   │              │
│  └────┬─────┘  └────┬─────┘  └──────────┘              │
│       │              │                                   │
│       ▼              ▼                                   │
│  ┌──────────┐  ┌──────────┐                              │
│  │  Rules   │  │  Memory  │                              │
│  │  (20)    │  │ (PgSQL)  │                              │
│  └──────────┘  └──────────┘                              │
│                                                         │
│  ┌──────────────────────────────────────┐                │
│  │  Self-Learning Pipeline             │                │
│  │  instincts → consolidate → rules    │                │
│  └──────────────────────────────────────┘                │
│                                                         │
│  ┌──────────────────────────────────────┐                │
│  │  Canavar Cross-Training             │                │
│  │  error-ledger → skill-matrix → team │                │
│  └──────────────────────────────────────┘                │
└─────────────────────────────────────────────────────────┘

Agent 分类

类别	数量	示例
Core Dev	14	frontend-dev, backend-dev, kraken, spark, devops, website-cloner
Review & QA	8	code-reviewer, security-reviewer, verifier, qa-engineer
Domain Experts	35	graphql-expert, kubernetes-expert, ddd-expert, redis-expert
Architecture	8	architect, planner, clean-arch-expert, cqrs-expert
Testing	6	tdd-guide, e2e-runner, arbiter, mocksmith
DevOps & Cloud	12	aws-expert, gcp-expert, azure-expert, terraform-expert
Analysis	11	scout, sleuth, data-analyst, profiler, strategist
Orchestration	16	nexus, sentinel, commander, neuron, vault, nitro
Documentation	6	technical-writer, doc-updater, copywriter, api-doc-generator, document-generator
Learning	7	self-learner, canavar, reputation-engine, session-replay-analyzer

技术栈

组件	技术
Runtime	Claude Code (Claude Max)
Models	Opus 4.6 / Sonnet 4.6
Hook engine	TypeScript → esbuild → .mjs
Memory DB	PostgreSQL + pgvector (Docker)
Agent format	Markdown + YAML frontmatter
Skill format	prompt.md / SKILL.md
Cross-training	JSONL ledger + JSON skill matrix

设计哲学

hooks are sensors. observe, filter, signal.
agents are muscles. build, produce, fix.
the bridge between them: context injection.
no direct RPC. no message passing. by design.
implicit coordination through context.

---

License

MIT

---