mcp-efficiency-engine 0.1.0

package/observability/logs/session/repomix-refresh-20260705-115419.log

@@ -0,0 +1,178 @@
+
+📦 Repomix v1.16.0
+
+
+[?25lΓáï Initializing...
+
+
+[?25lΓáÖ Searching for files...
+
+
+[?25l⠹ Searching for files...
+
+
+[?25l⠸ Searching for files...
+
+
+[?25l⠼ Searching for files...
+
+
+[?25l⠴ Searching for files...
+
+
+[?25l⠦ Searching for files...
+
+
+[?25l⠧ Searching for files...
+
+
+[?25lΓáç Searching for files...
+
+
+[?25l⠏ Searching for files...
+
+
+[?25lΓáï Searching for files...
+
+
+[?25lΓáÖ Collect file... (97/272) context/graphify-out/cache/ast/v0.9.2/3df5512333e56cb8ac86720b06992a52e73e8dd5b0ba13cd5e08474b99f9f406.json
+
+
+[?25l⠹ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠸ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠼ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠴ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠦ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠧ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25lΓáç Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠏ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25lΓáï Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25lΓáÖ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠹ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠸ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠼ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠴ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠦ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠧ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25lΓáç Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠏ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25lΓáï Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25lΓáÖ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠹ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠸ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠼ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠴ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠦ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠧ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25lΓáç Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠏ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25lΓáï Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25lΓáÖ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠹ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠸ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠼ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠴ Processing file... (272/272) tooling/tooling.manifest.json
+
+
+[?25l⠦ Running security check... (50/272) .github/pull_request_template.md
+
+
+[?25l⠧ Writing output file...
+
+
+[?25lΓáç Calculating metrics... (272/272) README.md
+
+
+[?25lΓ£ö Packing completed successfully!
+[?25h
+📈 Top 5 Files by Token Count:
+ΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇ
+1.  context/graphify-out/graph.json (25.424 tokens, 86.359 chars, 7.8%)
+2.  graphify-out/converted/Documento_funcional_portal_rcc_canarias_v1.05_1d4c5173.md (15.624 tokens, 58.959 chars, 4.8%)
+3.  docs/00-Ahorro_Tokens.md (13.444 tokens, 41.398 chars, 4.1%)
+4.  scripts/ops/hi.ps1 (13.000 tokens, 43.200 chars, 4%)
+5.  scripts/ops/compare-demo-results.ps1 (12.148 tokens, 40.683 chars, 3.7%)
+
+🔎 Security Check:
+ΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇ
+Γ£ö No suspicious files detected.
+
+
+📊 Pack Summary:
+ΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇΓöÇ
+  Total Files: 272 files
+ Total Tokens: 327.095 tokens
+  Total Chars: 1.126.552 chars
+       Output: context\repomix\repomix-output.xml
+     Security: Γ£ö No suspicious files detected
+
+🎉 All Done!
+Your repository has been successfully packed.
+
+💡 Repomix is now available in your browser! Try it at https://repomix.com

package/observability/logs/session/routing-evals-20260701-004221.log

@@ -0,0 +1,2 @@
+Routing evals completed: 5/5 passed
+Report: C:\repo\mcp-efficiency-engine\observability\evals\routing-eval-report.json

package/observability/logs/session/routing-evals-20260701-004235.log

@@ -0,0 +1,2 @@
+Routing evals completed: 5/5 passed
+Report: C:\repo\mcp-efficiency-engine\observability\evals\routing-eval-report.json

package/observability/logs/session/routing-evals-20260701-005629.log

@@ -0,0 +1,2 @@
+Routing evals completed: 5/5 passed
+Report: C:\repo\mcp-efficiency-engine\observability\evals\routing-eval-report.json

package/observability/logs/session/routing-evals-20260701-005834.log

@@ -0,0 +1,2 @@
+Routing evals completed: 5/5 passed
+Report: C:\repo\mcp-efficiency-engine\observability\evals\routing-eval-report.json

package/observability/logs/session/routing-evals-20260701-010437.log

@@ -0,0 +1,2 @@
+Routing evals completed: 5/5 passed
+Report: C:\repo\mcp-efficiency-engine\observability\evals\routing-eval-report.json

package/observability/logs/session/routing-evals-20260701-010603.log

@@ -0,0 +1,2 @@
+Routing evals completed: 5/5 passed
+Report: C:\repo\mcp-efficiency-engine\observability\evals\routing-eval-report.json

package/observability/logs/session/routing-evals-20260704-150125.log

@@ -0,0 +1,2 @@
+Routing evals completed: 8/8 passed
+Report: C:\repo\mcp-efficiency-engine\observability\evals\routing-eval-report.json

package/observability/logs/session/routing-evals-20260704-150157.log

@@ -0,0 +1,2 @@
+Routing evals completed: 8/8 passed
+Report: C:\repo\mcp-efficiency-engine\observability\evals\routing-eval-report.json

package/observability/logs/session/routing-evals-20260704-154434.log

@@ -0,0 +1,2 @@
+Routing evals completed: 8/8 passed
+Report: C:\repo\mcp-efficiency-engine\observability\evals\routing-eval-report.json

package/observability/logs/session/routing-evals-20260705-115327.log

@@ -0,0 +1,2 @@
+Routing evals completed: 8/8 passed
+Report: C:\repo\mcp-efficiency-engine\observability\evals\routing-eval-report.json

package/observability/logs/session/sibling-discovery-refresh-20260701-004250.log

@@ -0,0 +1,3 @@
+Dry-run mode. No registry changes applied.
+Proposal JSON: C:\repo\mcp-efficiency-engine\repo-intake\generated\reports\boost-discovery-proposal.json
+Proposal MD: C:\repo\mcp-efficiency-engine\repo-intake\generated\reports\boost-discovery-proposal.md

package/observability/logs/session/sibling-discovery-refresh-20260701-005850.log

@@ -0,0 +1,3 @@
+Dry-run mode. No registry changes applied.
+Proposal JSON: C:\repo\mcp-efficiency-engine\repo-intake\generated\reports\boost-discovery-proposal.json
+Proposal MD: C:\repo\mcp-efficiency-engine\repo-intake\generated\reports\boost-discovery-proposal.md

package/observability/logs/session/sibling-discovery-refresh-20260701-010618.log

@@ -0,0 +1,3 @@
+Dry-run mode. No registry changes applied.
+Proposal JSON: C:\repo\mcp-efficiency-engine\repo-intake\generated\reports\boost-discovery-proposal.json
+Proposal MD: C:\repo\mcp-efficiency-engine\repo-intake\generated\reports\boost-discovery-proposal.md

package/observability/logs/session/sibling-discovery-refresh-20260704-150238.log

@@ -0,0 +1,3 @@
+Dry-run mode. No registry changes applied.
+Proposal JSON: C:\repo\mcp-efficiency-engine\repo-intake\generated\reports\boost-discovery-proposal.json
+Proposal MD: C:\repo\mcp-efficiency-engine\repo-intake\generated\reports\boost-discovery-proposal.md

package/observability/logs/session/sibling-discovery-refresh-20260705-115418.log

@@ -0,0 +1,3 @@
+Dry-run mode. No registry changes applied.
+Proposal JSON: C:\repo\mcp-efficiency-engine\repo-intake\generated\reports\boost-discovery-proposal.json
+Proposal MD: C:\repo\mcp-efficiency-engine\repo-intake\generated\reports\boost-discovery-proposal.md

package/observability/logs.schema.json

@@ -0,0 +1,95 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://mcp-first-enterprise/observability/logs.schema.json",
+  "title": "Routing Decision Event",
+  "type": "object",
+  "required": [
+    "timestamp",
+    "input",
+    "intent",
+    "source_type",
+    "agent",
+    "engine",
+    "optimization_profile",
+    "fallback",
+    "grounded",
+    "sources"
+  ],
+  "properties": {
+    "event_id": { "type": "string", "minLength": 1 },
+    "timestamp": { "type": "string", "format": "date-time" },
+    "input": { "type": "string", "minLength": 1 },
+    "intent": { "type": "string", "minLength": 1 },
+    "source_type": { "type": "string", "minLength": 1 },
+    "agent": { "type": "string", "minLength": 1 },
+    "engine": { "type": "string", "minLength": 1 },
+    "optimization_profile": { "type": "string", "minLength": 1 },
+    "fallback": { "type": "boolean" },
+    "grounded": { "type": "boolean" },
+    "sources": {
+      "type": "array",
+      "items": { "type": "string" }
+    },
+    "notes": { "type": "string" },
+    "optimization": {
+      "type": "object",
+      "properties": {
+        "token_saver": { "const": "always_on" },
+        "token_saver_profile": {
+          "type": "string",
+          "enum": ["strict", "balanced", "evidence-first", "didactic"]
+        },
+        "caveman": { "const": "always_on" },
+        "caveman_profile": {
+          "type": "string",
+          "enum": ["full", "lite", "evidence-first", "didactic-lite"]
+        },
+        "sources_preserved": { "type": "boolean" },
+        "context_reduction_strategy": {
+          "type": "string",
+          "enum": ["symbol", "node", "chunk", "manifest", "snapshot-scope"]
+        }
+      },
+      "additionalProperties": true
+    },
+    "memory": {
+      "type": "object",
+      "properties": {
+        "selected": {
+          "type": "array",
+          "items": { "type": "string" }
+        },
+        "reason": { "type": "string" }
+      },
+      "additionalProperties": true
+    },
+    "learning": {
+      "type": "object",
+      "properties": {
+        "used_pattern": { "type": "string" },
+        "success": { "type": ["boolean", "null"] },
+        "outcome_status": {
+          "type": "string",
+          "enum": ["pending", "confirmed"]
+        },
+        "fallback": { "type": "boolean" },
+        "confidence": { "type": "number", "minimum": 0, "maximum": 1 }
+      },
+      "additionalProperties": true
+    },
+    "hitl": {
+      "type": "object",
+      "properties": {
+        "mode": { "type": "string", "enum": ["always_on_auto"] },
+        "required": { "type": "boolean" },
+        "action": {
+          "type": "string",
+          "enum": ["none", "request_human_confirmation", "block_until_human_approval"]
+        },
+        "reason": { "type": "string" }
+      },
+      "additionalProperties": true
+    }
+  },
+  "additionalProperties": true
+}

package/observability/logs.schema.md

@@ -0,0 +1,26 @@
+Schema: timestamp, input, intent, source_type, agent, engine, optimization_profile, fallback, grounded, sources, notes.
+
+## Optimization fields
+
+```json
+{
+  "optimization": {
+    "token_saver": "always_on",
+    "token_saver_profile": "strict|balanced|evidence-first|didactic",
+    "caveman": "always_on",
+    "caveman_profile": "full|lite|evidence-first|didactic-lite",
+    "sources_preserved": true,
+    "context_reduction_strategy": "symbol|node|chunk|manifest|snapshot-scope"
+  },
+  "memory": {
+  "selected": ["code-memory", "enterprise-memory"],
+  "reason": "multi-domain query"
+},
+"learning": {
+  "used_pattern": "auth+sla",
+  "success": true,
+  "fallback": false,
+  "confidence": 0.84
+}
+}
+```

package/observability/metrics.md

@@ -0,0 +1,35 @@
+# Metrics
+
+Métricas: routing accuracy, tool usage, tool misuse, tool switching, grounding quality, gaps, token efficiency, caveman effectiveness.
+
+## Optimization Metrics
+
+### Token Saver effectiveness
+
+- reduced_context_used: true/false
+- full_file_reads_avoided
+- chunks_retrieved
+- tool_calls_count
+- fallback_used
+
+### Caveman effectiveness
+
+- response_mode: lite/full/evidence-first/didactic-lite
+- unnecessary_intro_removed: true/false
+- sources_preserved: true/false
+- user_requested_more_detail: true/false
+
+## Umbrales operativos recomendados
+
+- routing_accuracy: `>= 95%`
+- grounding_quality (casos con fuentes): `>= 90%`
+- tool_misuse: `<= 2%`
+- fallback_rate: `<= 10%` (si supera, revisar reglas)
+- token_efficiency: tendencia estable o mejorando
+- caveman_effectiveness: `>= 85%`
+
+## Semaforo
+
+- `OK`: todos los umbrales en rango.
+- `WARN`: 1-2 metricas fuera de rango.
+- `FAIL`: 3+ metricas fuera de rango o fallo repetido de grounding/HITL.

package/observability/routing-evaluator.md

@@ -0,0 +1,37 @@
+# Routing Evaluator
+
+Comparar routing real contra matriz esperada.
+
+## Ejecución
+
+```powershell
+python .\scripts\intake\run-routing-evals.py
+```
+
+Entradas:
+
+- `observability/evals/routing-eval-cases.json`
+
+Salidas:
+
+- `observability/evals/routing-eval-report.json`
+- `observability/logs/routing-decisions.jsonl`
+
+Objetivo:
+
+- Detectar desvíos entre intención/dominio y agente/motor resultante.
+- Medir fallback rate y consistencia de perfiles de optimización.
+
+## Checklist de revision
+
+1. `cases_failed = 0`.
+1. Ningun caso con `prompt.exists=false`.
+1. Casos de alto riesgo con `hitl.required=true`.
+1. `fallback_rate` bajo umbral definido en `metrics.md`.
+
+## Triage rapido si falla
+
+1. Revisar desvio en `observability/evals/routing-eval-report.json`.
+1. Localizar evento en `observability/logs/routing-decisions.jsonl`.
+1. Ajustar regla en `scripts/intake/resolve-routing.py`.
+1. Reejecutar evals y registrar aprendizaje.

package/observability/scoring-rubric.md

@@ -0,0 +1,38 @@
+# Scoring Rubric
+
+## Componentes y pesos
+
+- Routing score: 40%
+- Grounding score: 25%
+- Efficiency score: 20%
+- Caveman score: 15%
+
+## Definicion de cada score
+
+### Routing score
+
+- Coincidencia de `agent` esperado.
+- Coincidencia de `engine` esperado.
+- Aplicacion correcta de `prompt.selected`.
+
+### Grounding score
+
+- `grounded=true` cuando el caso exige fuentes.
+- `sources` no vacio en consultas documentales.
+
+### Efficiency score
+
+- Perfil Token Saver correcto.
+- Sin tool misuse ni switching innecesario.
+
+### Caveman score
+
+- Respuesta directa y sin relleno.
+- Conserva evidencia cuando la tarea la exige.
+
+## Umbrales finales
+
+- `>= 90`: Excelente
+- `80-89`: Aceptable
+- `70-79`: Mejorable
+- `< 70`: No apto

package/optimization/ALWAYS_ON_OPTIMIZATION.md

@@ -0,0 +1,98 @@
+# Always-On Optimization Policy
+
+## Objetivo
+
+Caveman Mode y Token Saver deben estar activos por defecto para todo el sistema, incluidos flujos con GitHub Copilot, agentes, prompts, skills y routing.
+
+```txt
+Token Saver = siempre activo en la capa de contexto / retrieval / tools
+Caveman     = siempre activo en la capa de interacción / respuesta
+```
+
+Memory MCP y Learning tambien son always-on:
+
+```txt
+Memory MCP = memoria operativa/persistente y reutilizacion de patrones
+Learning   = feedback post-ejecucion para mejora continua del routing
+```
+
+## Frontera de capas
+
+1. Capa Always-On (fija): Token Saver + Caveman + Memory MCP + Learning.
+1. Capa Routing (variable): eleccion de agente/motor por intencion, dominio y fuente.
+1. Capa Ejecucion: toolchain especifico por motor elegido.
+
+Los motores (CodeGraph, GitNexus, Graphify, Azure RAG Builder, Repomix) no se fijan en Always-On: se resuelven en routing.
+
+## Regla global
+
+```txt
+Siempre optimizar contexto.
+Siempre optimizar respuesta.
+Nunca eliminar evidencia necesaria.
+Nunca eliminar claridad necesaria.
+```
+
+## Excepciones permitidas
+
+Caveman puede relajarse a modo didáctico si el usuario pide explícitamente:
+
+- explicación larga,
+- formación,
+- documentación para terceros,
+- storytelling,
+- material comercial/formativo.
+
+Token Saver NO se desactiva. Solo puede cambiar de intensidad:
+
+- `strict` para debug/coding/CLI,
+- `balanced` para arquitectura/RAG,
+- `evidence-first` para Azure RAG con fuentes obligatorias,
+- `didactic` para formación, manteniendo contexto mínimo suficiente.
+
+## Aplicación por motor
+
+| Motor | Token Saver | Caveman |
+| --- | --- | --- |
+| CodeGraph | strict | full en debug/coding |
+| GitNexus | balanced/strict | lite/full según caso |
+| Graphify | balanced | lite salvo formación |
+| Azure RAG Builder | evidence-first | lite, fuentes obligatorias |
+| Repomix | strict scope | N/A o lite |
+
+## Regla para Copilot
+
+GitHub Copilot debe leer esta política desde:
+
+- `.github/copilot-instructions.md`
+- `.github/instructions/always-on-optimization.instructions.md`
+- `.github/skills/token-saver/SKILL.md`
+- `.github/skills/caveman-mode/SKILL.md`
+
+Comandos caveman gestionados como capacidades propias del repo:
+
+- `.github/skills/caveman/SKILL.md`
+- `.github/skills/caveman-help/SKILL.md`
+- `.github/skills/caveman-review/SKILL.md`
+- `.github/skills/caveman-commit/SKILL.md`
+- `.github/skills/caveman-stats/SKILL.md`
+- `.github/skills/caveman-compress/SKILL.md`
+- `.github/skills/cavecrew/SKILL.md`
+- `.github/prompts/caveman.prompt.md`
+- `.github/prompts/caveman-help.prompt.md`
+- `.github/prompts/caveman-review.prompt.md`
+- `.github/prompts/caveman-commit.prompt.md`
+- `.github/prompts/caveman-stats.prompt.md`
+- `.github/prompts/caveman-compress.prompt.md`
+- `.github/prompts/cavecrew.prompt.md`
+
+Si hay conflicto, prevalece:
+
+```txt
+1. Seguridad / fuentes / grounding
+2. Token Saver
+3. Memory MCP
+4. Learning
+5. Caveman
+6. Estilo específico del agente
+```

package/optimization/caveman-mode.md

@@ -0,0 +1,62 @@
+# Caveman Mode
+
+## Definición
+
+Caveman Mode es la política de interacción para reducir ruido y acelerar loops.
+
+```txt
+Caveman = UX / interaction layer
+```
+
+## Objetivo
+
+Reducir:
+
+- explicaciones largas,
+- introducciones innecesarias,
+- repeticiones,
+- salida decorativa,
+- tokens de respuesta.
+
+## Modo estándar
+
+```txt
+Problema -> causa -> acción -> validación
+```
+
+## Ejemplo
+
+Normal:
+
+```txt
+He revisado el sistema y parece que podría haber un problema relacionado con...
+```
+
+Caveman:
+
+```txt
+Causa: token expirado no refresca redirect.
+Fix: revisar AuthService.refreshToken().
+Validar: test expiry + redirect.
+```
+
+## Niveles
+
+### Lite
+
+Bullets cortos, frases completas.
+
+### Full
+
+Muy directo, sin introducciones.
+
+### Ultra
+
+Solo datos críticos. Útil para CLI/debug.
+
+## No usar Caveman Ultra cuando
+
+- la respuesta es para formación,
+- el usuario pide explicación didáctica,
+- hay que convencer a stakeholders,
+- hay que documentar decisiones.

package/optimization/examples/login-bug.md

@@ -0,0 +1,25 @@
+# Example: Login Bug
+
+## Input
+
+Arregla login.
+
+## Ruta esperada
+
+- `agent`: `backend`
+- `engine`: `CodeGraph`
+- `token_saver_profile`: `strict`
+- `caveman_profile`: `full`
+
+## Validacion rapida
+
+```powershell
+py -3 .\scripts\intake\resolve-routing.py --input "Arregla bug de login" --intent bug-fix --domain dev --source-type code --capability code-fix
+```
+
+## Checks
+
+1. `fallback=false`
+1. `prompt.exists=true`
+1. `hitl.required=false` (si no hay accion de alto riesgo)
+