Review — mikeel_trainer

← overview

ventana últimos 7d vs 7d previos · tz America/Argentina/Buenos_Aires · último cambio Ninjo: 2026-05-06 18:09 (deploy config, hace 28d)
33
Nuevas (7d)
prior 23
1
Calls booked
50.0% de link
2
Link sent
3.1% de activas
$5
Revenue variable
150 msgs

Resumen — período vs período

Métricathis 7dprior 7dΔ
Activas6455+16.4%
Nuevas3323+43.5%
Link sent20
Calls booked11+0.0%
Booked/Link %50.0%
Booked/Activas %1.6%1.8%-0.3pp
Booked/Nuevas %3.0%4.3%-1.3pp
Revenue variable$5.25$4.66+12.7%

Funnel — semana actual vs anterior

PeríodoActivasNuevasLink sentCalls bookedBooked/LinkBooked/ActivasBooked/Nuevas
Semana actual (7d)64332150.0%1.6%3.0%
Semana anterior (7d)5523011.8%4.3%
Ver desglose por día
DíaActivasNuevasLink sentCalls bookedBooked/LinkBooked/ActivasBooked/Nuevas
06-0282000.0%0.0%
06-015311100.0%20.0%33.3%
05-3154100.0%0.0%0.0%
05-3052000.0%0.0%
05-291610000.0%0.0%
05-28219000.0%0.0%
05-2743000.0%0.0%
05-2674000.0%0.0%
05-25115000.0%0.0%
05-24108000.0%0.0%
05-2330000.0%
05-2232000.0%0.0%
05-21700114.3%
05-20144000.0%0.0%
run 2026-06-02T17:28:49+00:00 · fuente: live DB state (daily_audit pipeline)
8/10
BP semánticos PASS
top-10 best practices
12/16
Determinísticos PASS
2 FAIL · 2 WARN
C0 · H0 · M2
Fails por severidad
critical / high / medium
Drift
DB vs repo local

BP semánticos (top-10 best practices)

BP-1
BP-2
BP-3
BP-4
BP-5
BP-6
BP-7
BP-8
BP-9
BP-10
BPSevEstadoDetalle
BP-1HIGHPASSTemplate compliance
All 7 required sections for principles template are present in system.md: Identity + Mission (~100 words), Voice (6 real messages + summary), Flow (4 linear Etapas with inline guards), Calibration (energy states with actions), Principles (9 items, exceeds 5-7 but all actionable), Tools (7 tools listed with triggers), NO_RESPONSE (mandatory section present). Content fits template intent.
BP-2MEDIUMPASSExample coverage and cap (7 base + conditional, ≤20 total)
examples.md contains 18 examples covering all 7 base scenarios (happy path, resource delivery, objection handling, disqualification, emotional/sensitive, re-engagement, off-topic) plus 2 conditional scenarios (keyword trigger in Example 1, price objection in Example 8). Total count is 18, within the ≤20 cap.
BP-3HIGHPASSExamples align with the flow
All 18 examples follow the 4-Etapa flow (Entrada → Descubrimiento → Recurso → Programa + Llamada) in order. Each example reaches a terminal state: booking confirmed, NO_RESPONSE, cierre cálido, or manual_booking_requested. No example ends mid-flow.
BP-4CRITICALPASSNo internal contradictions in system.md
system.md contains one potential tension: Etapa 1 states 'No hay límite de intercambios — avanzar solo cuando el contexto es real, no por cantidad de mensajes' AND Etapa 3 requires 'al menos 2 preguntas de profundidad respondidas + 1 validación emocional hecha + 1 repregunta' before advancing. However, these are not contradictions — the first rule governs *when* to advance (quality-based, not message-count-based); the second rule specifies *what quality* is required. They apply under the same precondition (deciding to advance from Etapa 1→3) but are complementary, not conflicting. Examples 1, 10, 13, 14 demonstrate both rules coexisting correctly.
BP-5CRITICALPASSNo cross-file contradictions
All SDK files align on behavior and data. Keywords.json declares 5 keywords (INFO, PLAN, CAMBIO, DOCE, QUIERO) all with identical literal_response ('buenísimo! gracias por escribirme 🤗 comentame... qué te hizo tomar la decisión de hablarme?'), which matches system.md Etapa 0 opener. Examples 1, 3, 6, 8 demonstrate keyword triggers correctly. program.json pricing ($597 3mo, $997 6mo) is never mentioned in chat per system.md Principle 4 and Etapa 3 rule ('Nunca mandar link sin sí del lead'), and examples consistently avoid price disclosure. resources.json booking URL matches the Calendly link used in all examples. No file contradicts another under the same precondition.
BP-6MEDIUMPASSVoice consistency across examples
All 18 examples use consistent rioplatense voice: voseo (vos, tenés, querés), direct tone, short messages (1-2 sentences, 15-25 words), minimal emojis (🤗 💪 ✨ 🙌), no em-dashes or lists, natural prose. Tone varies appropriately (empathetic, curious, direct, warm) but the underlying persona is uniform. A reader would recognize all examples as the same agent.
BP-7CRITICALPASSUnguarded actions (PAT-004)
All risky actions in system.md have inline guards. Sending the booking link is guarded: 'Nunca mandar link sin sí del lead' (Etapa 3) and 'Invitar a llamada solo si hay contexto de calidad real: al menos 2 preguntas de profundidad respondidas + 1 validación emocional hecha + 1 repregunta usando las palabras del lead.' The guard is specific (not generic) and enumerates exact criteria. Examples 1, 10, 13, 14, 15 demonstrate the guard firing correctly when criteria are met. NO_RESPONSE is guarded by explicit trigger list. No unguarded action exists.
BP-8HIGHPASSNo stale version language in examples
All examples use current flow language from system.md. Openers match Etapa 0 ('buenísimo! gracias por escribirme 🤗 comentame... qué te hizo tomar la decisión de hablarme?'). Booking CTA matches Etapa 3 ('Perfecto! Preferís agendarlo vos directo con un link, o coordinamos por acá un día y un horario que te quede cómodo?'). NO_RESPONSE is literal token, not stale phrasing. No example uses outdated CTAs or flow steps.
BP-9MEDIUMFAILNo duplicated instructions
Principle 5 ('Diagnóstico de calidad antes de programa — No avanzar a programa o llamada hasta tener contexto real: mínimo 2 preguntas de profundidad respondidas (no superficiales), 1 validación emocional genuina, y 1 repregunta usando las palabras que usó el lead.') is restated verbatim in Etapa 3 ('Invitar a llamada solo si hay contexto de calidad real: al menos 2 preguntas de profundidad respondidas + 1 validación emocional hecha + 1 repregunta usando las palabras del lead.'). The same instruction appears in two locations with nearly identical wording. This is duplication, not a principle reinforced by an example.
evidence (2)
  • {'file': 'system.md', 'quote': "Principle 5: 'Diagnóstico de calidad antes de programa — No avanzar a programa o llamada hasta tener contexto real: mínimo 2 preguntas de profundidad respondidas'"}
  • {'file': 'system.md', 'quote': "Etapa 3: 'Invitar a llamada solo si hay contexto de calidad real: al menos 2 preguntas de profundidad respondidas + 1 validación emocional hecha'"}
BP-10MEDIUMFAILRule-vs-principle balance
Principles 1–3 and 8–9 are behavioral rules disguised as principles. Principle 1 ('No inventar') is a safety boundary, not behavior. Principle 2 ('No diagnosticar, no prometer resultados') is a safety boundary. Principle 3 ('Menores de edad → derivar') is a safety boundary. Principle 8 ('Precio temprano → redirigir a descubrimiento') is a hard rule about flow branching, not a principle. Principle 9 ('Max 1 caso de éxito por mensaje') is a formatting rule. These should be expressed as examples or moved to a separate Safety/Rules section. The template guidance states: 'Behavioral guidance (tone, flow, conversational style) should be expressed as principles + examples. Hard rules should remain only for factual constraints, safety boundaries, and compliance constraints.' Principles 1–3 are safety (acceptable), but 8–9 are behavioral rules that should be demonstrated via examples, not stated as principles.
evidence (2)
  • {'file': 'system.md', 'quote': 'Principle 8: \'Precio temprano → redirigir a descubrimiento — Si preguntan precio antes de diagnóstico: "el precio depende de tu situación, prefiero entender tu caso primero"\''}
  • {'file': 'system.md', 'quote': "Principle 9: 'Max 1 caso de éxito por mensaje — No amontonar testimonios. Uno alcanza para generar confianza.'"}

Checks determinísticos

6a
6c
6d
6e
6g
6h
6i
6j
6k
6l
6m
6n
6q
6r
6s
6t
CheckSecEstadoNotas
6a1.8PASSNO_RESPONSE present (limits not checked, --no-db)
6c3.xPASS104 lines
6d1.2PASS18 examples
6e1.3N/Ano Principles section
6gPAT-006WARNnon-STEP-0 keyword handling
6hAnti-patPASSno 'burbuja'
6iAnti-patPASSno HTML comments
6jAnti-patFAIL62 em-dashes across 1 file(s); first: system.md:1
6kPAT-006PASSno tool syntax leak
6lResourcesPASSno placeholder URLs
6mHygienePASSno audit metadata in prompt files
6nPAT-008PASSURL forms consistent across files
6qPAT-001PASSno instruction duplication into examples.md, keywords.json, program.json, resources.json
6rKeyword-driftFAIL2 in keywords.json + recent digest, never named in system.md: ['DOCE', 'QUIERO']
6sHygieneWARN1 expired date reference(s) in SDK files
6tCoverageN/Ano Principles or Flujo section found in system.md
canal ninjō-operator-mikel (C0AN999EZMK) · sync 2026-06-02
47
Digests
10 en últimos 14d
1
Cambios pedidos / pendientes
del checklist
49
Threads
en el canal
2026-06-02
Última sync
del digest

Cosas nuevas (digests recientes)

FechaTemaTags
2026-06-01weekly funnel metrics cambio keyword validationmetrics, keyword-performance, funnel-analysis, content-strategy
2026-06-01buyer attribution tracking pendingattribution, data-tracking, buyer-mapping, unresolved
2026-06-01weekly funnel report cambio keyword performancereporting, keyword-performance, conversion-analysis, content-strategy
2026-06-01buyer attribution loop cortex request pendingbuyer-attribution, data-request, unresolved, pending-response
2026-05-31high value lead intervention maritasilvettilead-scoring, manual-intervention, booking-pending, warm-lead
2026-05-31high value lead maritasilvetti manual interventionlead-scoring, manual-intervention, booking-pending, high-priority
2026-05-26program quota status checkprogram-management, capacity-planning
2026-05-26program capacity inquirycapacity, program, pending-response
2026-05-25weekly funnel report may 18 24 volume declinemetrics, funnel, weekly-report, volume-concern
2026-05-25weekly report may 18 24 funnel performancemetrics, funnel, weekly-report, keywords
2026-05-19whatsapp precall confirmation agent deploymentagent-deployment, whatsapp, booking-flow, precall
2026-05-18buyer attribution loop blockerattribution, analytics, blocker, recurring-request
2026-05-18weekly metrics may 11 17 funnel analysismetrics, funnel, weekly-report, keywords, conversion
2026-05-15high value lead intervention daianfuenteslead-intervention, high-value, manual-close, booking-stage
2026-05-13pre call message template standardizedbooking-confirmation, template, communication
2026-05-13pre call message template standardizationtemplates, booking-flow, pre-call
2026-05-11program quota opening confirmationprogram-launch, sales-readiness
2026-05-11quota opening sales cycle launchsales, quota-opening, program-launch
2026-05-11weekly funnel report may 4 10metrics, funnel, keywords, cambio-keyword
2026-05-11weekly metrics may 4 10 volume drop keyword analysismetrics, funnel, keyword-performance, action-items

Cambios pedidos por el creador / pendientes

IDEstadoCambio pedido
D5pendientePost-keyword `¡sí, mándamelo!`: el lead ya dijo "sí" — eliminar pasos de calificación, ir directo al link
sentiment del cliente → Ninjo · caché (2026-06-03 01:12:04) · modelo claude-sonnet-4-6 · ventana 7d
+1
Sentiment score
conforme
Bajo
Churn risk
del lado del cliente
1
Cambios pedidos
pendientes en checklist
6
Volumen cliente
agente: 33 convs

Matriz de volumen

enganchado sano
Cliente activo con tono ok — buena señal de retención.

Sentiment del cliente

📋 Cliente operativamente comprometido y sin señales de queja hacia Ninjo; la única fricción detectable es la no-respuesta reiterada a solicitudes de datos de atribución (buyer handles), que bloquea optimización interna pero no indica insatisfacción con la plataforma. Acción CS: escalar a Mikel en próxima touchpoint la importancia de proveer los buyer handles — enmarcarla como acción de alto impacto para él (mejora ROI del keyword `cambio`), no como deuda técnica de Ninjo.
SeñalValorEvidencia (cita del cliente)
Enojado con la plataformano
Es quejano
Queja constructiva (accionable)no

Evidencia citada

SeñalCita textualFecha
Mikel aprobó entrega de métricas de abril sin objeción“Mikel approved April metrics delivery (2,235 inbound, 149 outbound, 12 booked)”2026-05-04
Mikel proveyó datos de buyers de abril para análisis de atribución — señal de colaboración activa“Provided 5 April buyer names for attribution analysis (Antonella Ojeda, Ariana Aranda, Maria Florencia Sanchez, Candela Simon, Maria Jesus Vullez)”2026-04-30
Mikel rechazó recomendaciones de setters — criterio propio, no frustración con la plataforma“Declined setter recommendations; flow additions do not warrant implementation at current booking pace”2026-04-28
No responde solicitudes repetidas de buyer attribution (3 veces pedido sin respuesta) — fricción operativa por omisión, no queja explícita“Cortex pide consistentemente a Mikel handles de buyers para cruzar keyword → contenido → buyer (3 veces pedido, sin respuesta)”

Cambios pedidos por el creador (Métrica 1)

Un creador que pide cambios concretos está enganchado (mejor señal que el silencio).

IDEstadoCambio pedido
D5pendientePost-keyword `¡sí, mándamelo!`: el lead ya dijo "sí" — eliminar pasos de calificación, ir directo al link