@jaguilar87/gaia-ops 2.6.0 → 2.6.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +2 -2
- package/agents/README.en.md +289 -0
- package/agents/README.md +289 -0
- package/bin/README.en.md +187 -0
- package/bin/README.md +371 -0
- package/bin/gaia-cleanup.js +43 -0
- package/bin/gaia-metrics.js +314 -0
- package/bin/gaia-update.js +6 -4
- package/commands/README.en.md +385 -0
- package/commands/README.md +385 -0
- package/config/README.en.md +191 -0
- package/config/README.md +439 -0
- package/config/documentation-principles.en.md +341 -0
- package/config/documentation-principles.md +341 -0
- package/hooks/README.en.md +188 -0
- package/hooks/README.md +361 -0
- package/package.json +3 -2
- package/templates/README.en.md +141 -0
- package/templates/README.md +259 -0
- package/templates/settings.template.json +96 -585
- package/config/permission-rules-executive-summary.md +0 -277
- package/speckit/templates/tasks-template-bkp.md +0 -136
- package/tests/permissions-validation/empirical-permission-testing.md +0 -233
package/README.md
CHANGED
|
@@ -194,7 +194,7 @@ Este paquete sigue [Versionamiento Semántico](https://semver.org/):
|
|
|
194
194
|
- **MINOR:** Nuevas características, agentes o mejoras
|
|
195
195
|
- **PATCH:** Correcciones de bugs, clarificaciones, errores tipográficos
|
|
196
196
|
|
|
197
|
-
Versión actual: **2.2.
|
|
197
|
+
Versión actual: **2.2.4** (Actualizado)
|
|
198
198
|
|
|
199
199
|
Ver [CHANGELOG.md](./CHANGELOG.md) para el historial de versiones.
|
|
200
200
|
|
|
@@ -231,7 +231,7 @@ Ver [rnd-project-context](https://bitbucket.org/aaxisdigital/rnd-project-context
|
|
|
231
231
|
|
|
232
232
|
- **Issues:** [GitHub Issues](https://github.com/metraton/gaia-ops/issues)
|
|
233
233
|
- **Repositorio:** [github.com/metraton/gaia-ops](https://github.com/metraton/gaia-ops)
|
|
234
|
-
- **Autor:** Jorge Aguilar <jaguilar1897@gmail.com>
|
|
234
|
+
- **Autor:** Jorge Aguilar <jaguilar1897@gmail.com> (Verificado por Claude Code)
|
|
235
235
|
|
|
236
236
|
## Licencia
|
|
237
237
|
|
|
@@ -0,0 +1,289 @@
|
|
|
1
|
+
# Gaia-Ops Specialist Agents
|
|
2
|
+
|
|
3
|
+
**[🇪🇸 Versión en español](README.md)**
|
|
4
|
+
|
|
5
|
+
Agents are AI specialists that handle specific tasks in your DevOps infrastructure. Each agent is an expert in a particular domain and works in coordination with the main orchestrator.
|
|
6
|
+
|
|
7
|
+
## 🎯 Purpose
|
|
8
|
+
|
|
9
|
+
Agents allow complex work to be divided into manageable specialties. Instead of having a single system that tries to do everything, each agent focuses on what they do best - like having a team of experts instead of a generalist.
|
|
10
|
+
|
|
11
|
+
**Problem it solves:** DevOps tasks are diverse and complex. A single agent cannot be an expert in everything (Terraform, Kubernetes, GCP, AWS, applications). Specialist agents enable depth of knowledge in each area.
|
|
12
|
+
|
|
13
|
+
## 🔄 How It Works
|
|
14
|
+
|
|
15
|
+
### Architecture Flow
|
|
16
|
+
|
|
17
|
+
```
|
|
18
|
+
User sends question
|
|
19
|
+
↓
|
|
20
|
+
[Orchestrator (CLAUDE.md)]
|
|
21
|
+
↓
|
|
22
|
+
[Agent Router] ← analyzes question
|
|
23
|
+
↓
|
|
24
|
+
Selects appropriate agent
|
|
25
|
+
↓
|
|
26
|
+
┌───┴───┬───────┬────────┬─────────┬────────┐
|
|
27
|
+
↓ ↓ ↓ ↓ ↓ ↓
|
|
28
|
+
[terraform] [gitops] [gcp] [aws] [devops] [gaia]
|
|
29
|
+
architect operator troubl. troubl. developer meta-agent
|
|
30
|
+
↓ ↓ ↓ ↓ ↓ ↓
|
|
31
|
+
└───┬───┴───────┴────────┴─────────┴────────┘
|
|
32
|
+
↓
|
|
33
|
+
[Context Provider] ← provides relevant information
|
|
34
|
+
↓
|
|
35
|
+
Agent executes task
|
|
36
|
+
↓
|
|
37
|
+
Result to user
|
|
38
|
+
```
|
|
39
|
+
|
|
40
|
+
### Real Example Flow
|
|
41
|
+
|
|
42
|
+
```
|
|
43
|
+
Example: "Deploy auth service to production cluster"
|
|
44
|
+
|
|
45
|
+
1. User asks question
|
|
46
|
+
↓
|
|
47
|
+
2. [Orchestrator] receives request
|
|
48
|
+
↓
|
|
49
|
+
3. [Agent Router] analyzes keywords:
|
|
50
|
+
- "deploy" → deployment operation
|
|
51
|
+
- "service" → Kubernetes application
|
|
52
|
+
- "cluster" → GitOps
|
|
53
|
+
↓
|
|
54
|
+
4. Router selects → **gitops-operator**
|
|
55
|
+
↓
|
|
56
|
+
5. [Context Provider] prepares information:
|
|
57
|
+
- Current namespace
|
|
58
|
+
- Existing releases
|
|
59
|
+
- Cluster configuration
|
|
60
|
+
↓
|
|
61
|
+
6. [gitops-operator] receives context and question
|
|
62
|
+
↓
|
|
63
|
+
7. Agent generates plan:
|
|
64
|
+
- Update deployment.yaml
|
|
65
|
+
- Increment image version
|
|
66
|
+
- Apply with kubectl
|
|
67
|
+
↓
|
|
68
|
+
8. [Approval Gate] requests confirmation (T3 operation)
|
|
69
|
+
- Shows proposed changes
|
|
70
|
+
- User approves ✅
|
|
71
|
+
↓
|
|
72
|
+
9. [gitops-operator] executes:
|
|
73
|
+
- kubectl apply -f deployment.yaml
|
|
74
|
+
- kubectl rollout status deployment/auth
|
|
75
|
+
↓
|
|
76
|
+
10. Verifies success:
|
|
77
|
+
- Pods running: 3/3
|
|
78
|
+
- Health checks: OK
|
|
79
|
+
↓
|
|
80
|
+
11. Reports result: "✅ auth deployed successfully to production"
|
|
81
|
+
```
|
|
82
|
+
|
|
83
|
+
## 📋 Available Agents
|
|
84
|
+
|
|
85
|
+
### 1. terraform-architect 🏗️
|
|
86
|
+
**Expert in:** Infrastructure as code
|
|
87
|
+
|
|
88
|
+
Handles everything related to Terraform and Terragrunt. Like the architect who designs and builds the foundations of your cloud infrastructure.
|
|
89
|
+
|
|
90
|
+
**When to use:**
|
|
91
|
+
- Create GKE clusters
|
|
92
|
+
- Configure VPCs and networks
|
|
93
|
+
- Manage storage buckets
|
|
94
|
+
- Configure IAM permissions
|
|
95
|
+
|
|
96
|
+
**Example questions:**
|
|
97
|
+
- "Create a new GKE cluster for staging environment"
|
|
98
|
+
- "Add an additional subnet in us-east1"
|
|
99
|
+
|
|
100
|
+
**Tiers:** T0 (read), T1 (validate), T2 (plan), T3 (apply)
|
|
101
|
+
|
|
102
|
+
---
|
|
103
|
+
|
|
104
|
+
### 2. gitops-operator ⚙️
|
|
105
|
+
**Expert in:** Kubernetes and deployments
|
|
106
|
+
|
|
107
|
+
Handles applications in Kubernetes, deployments, services and everything related to GitOps. Like the operator who keeps applications running in clusters.
|
|
108
|
+
|
|
109
|
+
**When to use:**
|
|
110
|
+
- Deploy services
|
|
111
|
+
- Update deployments
|
|
112
|
+
- Configure ingress
|
|
113
|
+
- Scale applications
|
|
114
|
+
|
|
115
|
+
**Example questions:**
|
|
116
|
+
- "Deploy version 1.2.3 of the backend"
|
|
117
|
+
- "Scale auth service to 5 replicas"
|
|
118
|
+
|
|
119
|
+
**Tiers:** T0 (read), T1 (validate), T2 (plan), T3 (apply)
|
|
120
|
+
|
|
121
|
+
---
|
|
122
|
+
|
|
123
|
+
### 3. gcp-troubleshooter 🔍
|
|
124
|
+
**Expert in:** Google Cloud Platform diagnostics
|
|
125
|
+
|
|
126
|
+
Identifies problems and gathers information about GCP resources. Like the detective who investigates what's happening in the cloud.
|
|
127
|
+
|
|
128
|
+
**When to use:**
|
|
129
|
+
- Diagnose GCP errors
|
|
130
|
+
- Review Cloud Logging logs
|
|
131
|
+
- Check resource status
|
|
132
|
+
- Analyze IAM permissions
|
|
133
|
+
|
|
134
|
+
**Example questions:**
|
|
135
|
+
- "Why is the cluster failing?"
|
|
136
|
+
- "Show auth service logs from the last 2 hours"
|
|
137
|
+
|
|
138
|
+
**Tiers:** T0 only (read-only, makes no changes)
|
|
139
|
+
|
|
140
|
+
---
|
|
141
|
+
|
|
142
|
+
### 4. aws-troubleshooter 🔍
|
|
143
|
+
**Expert in:** Amazon Web Services diagnostics
|
|
144
|
+
|
|
145
|
+
Similar to gcp-troubleshooter but for AWS. Diagnoses problems and gathers information about AWS resources.
|
|
146
|
+
|
|
147
|
+
**When to use:**
|
|
148
|
+
- Diagnose AWS errors
|
|
149
|
+
- Review CloudWatch logs
|
|
150
|
+
- Check EC2/EKS resource status
|
|
151
|
+
- Analyze IAM policies
|
|
152
|
+
|
|
153
|
+
**Example questions:**
|
|
154
|
+
- "Why is the EKS cluster failing?"
|
|
155
|
+
- "Show EC2 instance metrics"
|
|
156
|
+
|
|
157
|
+
**Tiers:** T0 only (read-only)
|
|
158
|
+
|
|
159
|
+
---
|
|
160
|
+
|
|
161
|
+
### 5. devops-developer 💻
|
|
162
|
+
**Expert in:** Application code and CI/CD
|
|
163
|
+
|
|
164
|
+
Works with application code, Dockerfiles, builds and tests. Like the developer who ensures code works correctly.
|
|
165
|
+
|
|
166
|
+
**When to use:**
|
|
167
|
+
- Create/modify Dockerfiles
|
|
168
|
+
- Configure npm/yarn builds
|
|
169
|
+
- Write automation scripts
|
|
170
|
+
- Configure CI pipelines
|
|
171
|
+
|
|
172
|
+
**Example questions:**
|
|
173
|
+
- "Optimize the backend Dockerfile"
|
|
174
|
+
- "Add unit tests to the service"
|
|
175
|
+
|
|
176
|
+
**Tiers:** T0 (read), T1 (validate), T2 (test builds)
|
|
177
|
+
|
|
178
|
+
---
|
|
179
|
+
|
|
180
|
+
### 6. Gaia 🧠
|
|
181
|
+
**Expert in:** The agent system itself
|
|
182
|
+
|
|
183
|
+
The meta-agent that understands how the entire orchestration system works. Like the systems architect who optimizes and documents the operation of the agents themselves.
|
|
184
|
+
|
|
185
|
+
**When to use:**
|
|
186
|
+
- Analyze system logs
|
|
187
|
+
- Optimize agent routing
|
|
188
|
+
- Improve documentation
|
|
189
|
+
- Diagnose orchestrator problems
|
|
190
|
+
|
|
191
|
+
**Example questions:**
|
|
192
|
+
- "Why did routing fail in this case?"
|
|
193
|
+
- "Analyze agent router accuracy"
|
|
194
|
+
|
|
195
|
+
**Tiers:** T0-T2 (analysis and proposals, doesn't execute changes)
|
|
196
|
+
|
|
197
|
+
## 🚀 How Agents Are Invoked
|
|
198
|
+
|
|
199
|
+
### Automatic Invocation (Recommended)
|
|
200
|
+
|
|
201
|
+
The orchestrator analyzes your question and automatically selects the appropriate agent:
|
|
202
|
+
|
|
203
|
+
```bash
|
|
204
|
+
# In Claude Code, simply ask:
|
|
205
|
+
"Deploy auth-service version 1.2.3"
|
|
206
|
+
# → Orchestrator automatically invokes gitops-operator
|
|
207
|
+
```
|
|
208
|
+
|
|
209
|
+
### Manual Invocation (Advanced)
|
|
210
|
+
|
|
211
|
+
For specific cases where you want to directly invoke an agent:
|
|
212
|
+
|
|
213
|
+
```bash
|
|
214
|
+
# Use the Task command
|
|
215
|
+
Task(
|
|
216
|
+
subagent_type="gitops-operator",
|
|
217
|
+
description="Deploy auth service",
|
|
218
|
+
prompt="Deploy auth-service version 1.2.3 to production cluster"
|
|
219
|
+
)
|
|
220
|
+
```
|
|
221
|
+
|
|
222
|
+
## 🔧 Technical Details
|
|
223
|
+
|
|
224
|
+
### Agent Structure
|
|
225
|
+
|
|
226
|
+
Each agent is a Markdown file (`agent.md`) with these sections:
|
|
227
|
+
|
|
228
|
+
```markdown
|
|
229
|
+
---
|
|
230
|
+
name: agent-name
|
|
231
|
+
description: Brief description
|
|
232
|
+
tools: List of allowed tools
|
|
233
|
+
model: Model configuration
|
|
234
|
+
---
|
|
235
|
+
|
|
236
|
+
# Agent Name
|
|
237
|
+
|
|
238
|
+
[Comprehensive instructions for the agent]
|
|
239
|
+
```
|
|
240
|
+
|
|
241
|
+
### Security Tiers
|
|
242
|
+
|
|
243
|
+
Agents operate at different security levels:
|
|
244
|
+
|
|
245
|
+
| Tier | Description | Requires Approval |
|
|
246
|
+
|------|-------------|-------------------|
|
|
247
|
+
| **T0** | Read-only (get, describe, list) | No |
|
|
248
|
+
| **T1** | Validation (validate, dry-run, test) | No |
|
|
249
|
+
| **T2** | Planning (plan, simulate) | No |
|
|
250
|
+
| **T3** | Execution (apply, create, delete) | **Yes** ✅ |
|
|
251
|
+
|
|
252
|
+
**Important note:** T3 operations ALWAYS require explicit user approval through the Approval Gate.
|
|
253
|
+
|
|
254
|
+
### Smart Routing
|
|
255
|
+
|
|
256
|
+
The system uses multiple techniques to select the right agent:
|
|
257
|
+
|
|
258
|
+
1. **Keywords:** Domain-specific terms
|
|
259
|
+
2. **Semantic matching:** Semantic similarity using embeddings
|
|
260
|
+
3. **Context awareness:** Considers project context
|
|
261
|
+
|
|
262
|
+
**Current accuracy:** ~92.7% (based on tests)
|
|
263
|
+
|
|
264
|
+
## 📖 References
|
|
265
|
+
|
|
266
|
+
**Related documentation:**
|
|
267
|
+
- [Orchestration Workflow](../config/orchestration-workflow.md) - How a request flows
|
|
268
|
+
- [Agent Catalog](../config/agent-catalog.md) - Complete details of each agent
|
|
269
|
+
- [Context Contracts](../config/context-contracts.md) - What information each agent receives
|
|
270
|
+
- [Agent Router](../tools/1-routing/agent_router.py) - Routing code
|
|
271
|
+
|
|
272
|
+
**Agent files:**
|
|
273
|
+
```
|
|
274
|
+
agents/
|
|
275
|
+
├── terraform-architect.md (~800 lines)
|
|
276
|
+
├── gitops-operator.md (~750 lines)
|
|
277
|
+
├── gcp-troubleshooter.md (~600 lines)
|
|
278
|
+
├── aws-troubleshooter.md (~600 lines)
|
|
279
|
+
├── devops-developer.md (~500 lines)
|
|
280
|
+
└── gaia.md (~1650 lines)
|
|
281
|
+
```
|
|
282
|
+
|
|
283
|
+
---
|
|
284
|
+
|
|
285
|
+
**Version:** 1.0.0
|
|
286
|
+
**Last updated:** 2025-11-14
|
|
287
|
+
**Total agents:** 6 specialists
|
|
288
|
+
**Maintained by:** Gaia (meta-agent)
|
|
289
|
+
|
package/agents/README.md
ADDED
|
@@ -0,0 +1,289 @@
|
|
|
1
|
+
# Agentes Especialistas de Gaia-Ops
|
|
2
|
+
|
|
3
|
+
**[🇺🇸 English version](README.en.md)**
|
|
4
|
+
|
|
5
|
+
Los agentes son especialistas de IA que manejan tareas específicas en tu infraestructura DevOps. Cada agente es experto en un dominio particular y trabaja de forma coordinada con el orquestador principal.
|
|
6
|
+
|
|
7
|
+
## 🎯 Propósito
|
|
8
|
+
|
|
9
|
+
Los agentes permiten dividir el trabajo complejo en especialidades manejables. En lugar de tener un solo sistema que intente hacerlo todo, cada agente se enfoca en lo que mejor sabe hacer - como tener un equipo de expertos en lugar de un generalista.
|
|
10
|
+
|
|
11
|
+
**Problema que resuelve:** Las tareas DevOps son diversas y complejas. Un agente único no puede ser experto en todo (Terraform, Kubernetes, GCP, AWS, aplicaciones). Los agentes especialistas permiten profundidad de conocimiento en cada área.
|
|
12
|
+
|
|
13
|
+
## 🔄 Cómo Funciona
|
|
14
|
+
|
|
15
|
+
### Flujo de Arquitectura
|
|
16
|
+
|
|
17
|
+
```
|
|
18
|
+
Usuario envía pregunta
|
|
19
|
+
↓
|
|
20
|
+
[Orquestador (CLAUDE.md)]
|
|
21
|
+
↓
|
|
22
|
+
[Agent Router] ← analiza la pregunta
|
|
23
|
+
↓
|
|
24
|
+
Selecciona agente apropiado
|
|
25
|
+
↓
|
|
26
|
+
┌───┴───┬───────┬────────┬─────────┬────────┐
|
|
27
|
+
↓ ↓ ↓ ↓ ↓ ↓
|
|
28
|
+
[terraform] [gitops] [gcp] [aws] [devops] [gaia]
|
|
29
|
+
architect operator troubl. troubl. developer meta-agent
|
|
30
|
+
↓ ↓ ↓ ↓ ↓ ↓
|
|
31
|
+
└───┬───┴───────┴────────┴─────────┴────────┘
|
|
32
|
+
↓
|
|
33
|
+
[Context Provider] ← proporciona información relevante
|
|
34
|
+
↓
|
|
35
|
+
Agente ejecuta tarea
|
|
36
|
+
↓
|
|
37
|
+
Resultado al usuario
|
|
38
|
+
```
|
|
39
|
+
|
|
40
|
+
### Flujo de Ejemplo Real
|
|
41
|
+
|
|
42
|
+
```
|
|
43
|
+
Ejemplo: "Despliega el servicio auth en el cluster de producción"
|
|
44
|
+
|
|
45
|
+
1. Usuario hace la pregunta
|
|
46
|
+
↓
|
|
47
|
+
2. [Orquestador] recibe la solicitud
|
|
48
|
+
↓
|
|
49
|
+
3. [Agent Router] analiza palabras clave:
|
|
50
|
+
- "despliega" → operación de deployment
|
|
51
|
+
- "servicio" → aplicación en Kubernetes
|
|
52
|
+
- "cluster" → GitOps
|
|
53
|
+
↓
|
|
54
|
+
4. Router selecciona → **gitops-operator**
|
|
55
|
+
↓
|
|
56
|
+
5. [Context Provider] prepara información:
|
|
57
|
+
- Namespace actual
|
|
58
|
+
- Releases existentes
|
|
59
|
+
- Configuración del cluster
|
|
60
|
+
↓
|
|
61
|
+
6. [gitops-operator] recibe contexto y pregunta
|
|
62
|
+
↓
|
|
63
|
+
7. Agente genera plan:
|
|
64
|
+
- Actualizar deployment.yaml
|
|
65
|
+
- Incrementar versión de imagen
|
|
66
|
+
- Aplicar con kubectl
|
|
67
|
+
↓
|
|
68
|
+
8. [Approval Gate] pide confirmación (es operación T3)
|
|
69
|
+
- Muestra cambios propuestos
|
|
70
|
+
- Usuario aprueba ✅
|
|
71
|
+
↓
|
|
72
|
+
9. [gitops-operator] ejecuta:
|
|
73
|
+
- kubectl apply -f deployment.yaml
|
|
74
|
+
- kubectl rollout status deployment/auth
|
|
75
|
+
↓
|
|
76
|
+
10. Verifica éxito:
|
|
77
|
+
- Pods running: 3/3
|
|
78
|
+
- Health checks: OK
|
|
79
|
+
↓
|
|
80
|
+
11. Reporta resultado: "✅ auth desplegado exitosamente en producción"
|
|
81
|
+
```
|
|
82
|
+
|
|
83
|
+
## 📋 Agentes Disponibles
|
|
84
|
+
|
|
85
|
+
### 1. terraform-architect 🏗️
|
|
86
|
+
**Experto en:** Infraestructura como código
|
|
87
|
+
|
|
88
|
+
Maneja todo lo relacionado con Terraform y Terragrunt. Es como el arquitecto que diseña y construye los cimientos de tu infraestructura cloud.
|
|
89
|
+
|
|
90
|
+
**Cuándo se usa:**
|
|
91
|
+
- Crear clusters GKE
|
|
92
|
+
- Configurar VPCs y redes
|
|
93
|
+
- Gestionar buckets de almacenamiento
|
|
94
|
+
- Configurar permisos IAM
|
|
95
|
+
|
|
96
|
+
**Ejemplo de pregunta:**
|
|
97
|
+
- "Crea un nuevo cluster GKE para el ambiente de staging"
|
|
98
|
+
- "Agrega una subnet adicional en us-east1"
|
|
99
|
+
|
|
100
|
+
**Tiers:** T0 (leer), T1 (validar), T2 (planear), T3 (aplicar)
|
|
101
|
+
|
|
102
|
+
---
|
|
103
|
+
|
|
104
|
+
### 2. gitops-operator ⚙️
|
|
105
|
+
**Experto en:** Kubernetes y despliegues
|
|
106
|
+
|
|
107
|
+
Maneja aplicaciones en Kubernetes, deployments, services y todo lo relacionado con GitOps. Es como el operador que mantiene las aplicaciones funcionando en los clusters.
|
|
108
|
+
|
|
109
|
+
**Cuándo se usa:**
|
|
110
|
+
- Desplegar servicios
|
|
111
|
+
- Actualizar deployments
|
|
112
|
+
- Configurar ingress
|
|
113
|
+
- Escalar aplicaciones
|
|
114
|
+
|
|
115
|
+
**Ejemplo de pregunta:**
|
|
116
|
+
- "Despliega la versión 1.2.3 del backend"
|
|
117
|
+
- "Escala el servicio auth a 5 réplicas"
|
|
118
|
+
|
|
119
|
+
**Tiers:** T0 (leer), T1 (validar), T2 (planear), T3 (aplicar)
|
|
120
|
+
|
|
121
|
+
---
|
|
122
|
+
|
|
123
|
+
### 3. gcp-troubleshooter 🔍
|
|
124
|
+
**Experto en:** Diagnóstico de Google Cloud Platform
|
|
125
|
+
|
|
126
|
+
Identifica problemas y recopila información sobre recursos en GCP. Es como el detective que investiga qué está pasando en la nube.
|
|
127
|
+
|
|
128
|
+
**Cuándo se usa:**
|
|
129
|
+
- Diagnosticar errores en GCP
|
|
130
|
+
- Revisar logs de Cloud Logging
|
|
131
|
+
- Verificar estado de recursos
|
|
132
|
+
- Analizar permisos IAM
|
|
133
|
+
|
|
134
|
+
**Ejemplo de pregunta:**
|
|
135
|
+
- "¿Por qué está fallando el cluster?"
|
|
136
|
+
- "Muestra los logs del servicio auth en las últimas 2 horas"
|
|
137
|
+
|
|
138
|
+
**Tiers:** T0 únicamente (solo lectura, no hace cambios)
|
|
139
|
+
|
|
140
|
+
---
|
|
141
|
+
|
|
142
|
+
### 4. aws-troubleshooter 🔍
|
|
143
|
+
**Experto en:** Diagnóstico de Amazon Web Services
|
|
144
|
+
|
|
145
|
+
Similar a gcp-troubleshooter pero para AWS. Diagnostica problemas y recopila información sobre recursos en Amazon Web Services.
|
|
146
|
+
|
|
147
|
+
**Cuándo se usa:**
|
|
148
|
+
- Diagnosticar errores en AWS
|
|
149
|
+
- Revisar logs de CloudWatch
|
|
150
|
+
- Verificar estado de recursos EC2/EKS
|
|
151
|
+
- Analizar políticas IAM
|
|
152
|
+
|
|
153
|
+
**Ejemplo de pregunta:**
|
|
154
|
+
- "¿Por qué está fallando el EKS cluster?"
|
|
155
|
+
- "Muestra métricas de la instancia EC2"
|
|
156
|
+
|
|
157
|
+
**Tiers:** T0 únicamente (solo lectura)
|
|
158
|
+
|
|
159
|
+
---
|
|
160
|
+
|
|
161
|
+
### 5. devops-developer 💻
|
|
162
|
+
**Experto en:** Código de aplicaciones y CI/CD
|
|
163
|
+
|
|
164
|
+
Trabaja con código de aplicaciones, Dockerfiles, builds y tests. Es como el desarrollador que asegura que el código funcione correctamente.
|
|
165
|
+
|
|
166
|
+
**Cuándo se usa:**
|
|
167
|
+
- Crear/modificar Dockerfiles
|
|
168
|
+
- Configurar npm/yarn builds
|
|
169
|
+
- Escribir scripts de automatización
|
|
170
|
+
- Configurar CI pipelines
|
|
171
|
+
|
|
172
|
+
**Ejemplo de pregunta:**
|
|
173
|
+
- "Optimiza el Dockerfile del backend"
|
|
174
|
+
- "Agrega tests unitarios al servicio"
|
|
175
|
+
|
|
176
|
+
**Tiers:** T0 (leer), T1 (validar), T2 (probar builds)
|
|
177
|
+
|
|
178
|
+
---
|
|
179
|
+
|
|
180
|
+
### 6. Gaia 🧠
|
|
181
|
+
**Experto en:** El propio sistema de agentes
|
|
182
|
+
|
|
183
|
+
El meta-agente que entiende cómo funciona todo el sistema de orquestación. Es como el arquitecto de sistemas que optimiza y documenta el funcionamiento de los propios agentes.
|
|
184
|
+
|
|
185
|
+
**Cuándo se usa:**
|
|
186
|
+
- Analizar logs del sistema
|
|
187
|
+
- Optimizar routing de agentes
|
|
188
|
+
- Mejorar documentación
|
|
189
|
+
- Diagnosticar problemas del orquestador
|
|
190
|
+
|
|
191
|
+
**Ejemplo de pregunta:**
|
|
192
|
+
- "¿Por qué falló el routing en este caso?"
|
|
193
|
+
- "Analiza la precisión del agent router"
|
|
194
|
+
|
|
195
|
+
**Tiers:** T0-T2 (análisis y propuestas, no ejecuta cambios)
|
|
196
|
+
|
|
197
|
+
## 🚀 Cómo se Invocan los Agentes
|
|
198
|
+
|
|
199
|
+
### Invocación Automática (Recomendado)
|
|
200
|
+
|
|
201
|
+
El orquestador analiza tu pregunta y automáticamente selecciona el agente apropiado:
|
|
202
|
+
|
|
203
|
+
```bash
|
|
204
|
+
# En Claude Code, simplemente pregunta:
|
|
205
|
+
"Despliega auth-service versión 1.2.3"
|
|
206
|
+
# → El orquestador invoca gitops-operator automáticamente
|
|
207
|
+
```
|
|
208
|
+
|
|
209
|
+
### Invocación Manual (Avanzado)
|
|
210
|
+
|
|
211
|
+
Para casos específicos donde quieres invocar un agente directamente:
|
|
212
|
+
|
|
213
|
+
```bash
|
|
214
|
+
# Usar el comando Task
|
|
215
|
+
Task(
|
|
216
|
+
subagent_type="gitops-operator",
|
|
217
|
+
description="Deploy auth service",
|
|
218
|
+
prompt="Deploy auth-service version 1.2.3 to production cluster"
|
|
219
|
+
)
|
|
220
|
+
```
|
|
221
|
+
|
|
222
|
+
## 🔧 Características Técnicas
|
|
223
|
+
|
|
224
|
+
### Estructura de un Agente
|
|
225
|
+
|
|
226
|
+
Cada agente es un archivo Markdown (`agente.md`) con estas secciones:
|
|
227
|
+
|
|
228
|
+
```markdown
|
|
229
|
+
---
|
|
230
|
+
name: agent-name
|
|
231
|
+
description: Brief description
|
|
232
|
+
tools: List of allowed tools
|
|
233
|
+
model: Model configuration
|
|
234
|
+
---
|
|
235
|
+
|
|
236
|
+
# Agent Name
|
|
237
|
+
|
|
238
|
+
[Comprehensive instructions for the agent]
|
|
239
|
+
```
|
|
240
|
+
|
|
241
|
+
### Tiers de Seguridad
|
|
242
|
+
|
|
243
|
+
Los agentes operan en diferentes niveles de seguridad:
|
|
244
|
+
|
|
245
|
+
| Tier | Descripción | Requiere Aprobación |
|
|
246
|
+
|------|-------------|---------------------|
|
|
247
|
+
| **T0** | Solo lectura (get, describe, list) | No |
|
|
248
|
+
| **T1** | Validación (validate, dry-run, test) | No |
|
|
249
|
+
| **T2** | Planificación (plan, simulate) | No |
|
|
250
|
+
| **T3** | Ejecución (apply, create, delete) | **Sí** ✅ |
|
|
251
|
+
|
|
252
|
+
**Nota importante:** Las operaciones T3 SIEMPRE requieren aprobación explícita del usuario a través del Approval Gate.
|
|
253
|
+
|
|
254
|
+
### Routing Inteligente
|
|
255
|
+
|
|
256
|
+
El sistema usa múltiples técnicas para seleccionar el agente correcto:
|
|
257
|
+
|
|
258
|
+
1. **Palabras clave:** Términos específicos del dominio
|
|
259
|
+
2. **Semantic matching:** Similitud semántica usando embeddings
|
|
260
|
+
3. **Context awareness:** Considera el contexto del proyecto
|
|
261
|
+
|
|
262
|
+
**Precisión actual:** ~92.7% (basado en tests)
|
|
263
|
+
|
|
264
|
+
## 📖 Referencias
|
|
265
|
+
|
|
266
|
+
**Documentación relacionada:**
|
|
267
|
+
- [Orchestration Workflow](../config/orchestration-workflow.md) - Cómo fluye una solicitud
|
|
268
|
+
- [Agent Catalog](../config/agent-catalog.md) - Detalles completos de cada agente
|
|
269
|
+
- [Context Contracts](../config/context-contracts.md) - Qué información recibe cada agente
|
|
270
|
+
- [Agent Router](../tools/1-routing/agent_router.py) - Código del routing
|
|
271
|
+
|
|
272
|
+
**Archivos de agentes:**
|
|
273
|
+
```
|
|
274
|
+
agents/
|
|
275
|
+
├── terraform-architect.md (~800 líneas)
|
|
276
|
+
├── gitops-operator.md (~750 líneas)
|
|
277
|
+
├── gcp-troubleshooter.md (~600 líneas)
|
|
278
|
+
├── aws-troubleshooter.md (~600 líneas)
|
|
279
|
+
├── devops-developer.md (~500 líneas)
|
|
280
|
+
└── gaia.md (~1650 líneas)
|
|
281
|
+
```
|
|
282
|
+
|
|
283
|
+
---
|
|
284
|
+
|
|
285
|
+
**Versión:** 1.0.0
|
|
286
|
+
**Última actualización:** 2025-11-14
|
|
287
|
+
**Total de agentes:** 6 especialistas
|
|
288
|
+
**Mantenido por:** Gaia (meta-agent)
|
|
289
|
+
|