| Métrica | this 7d | prior 7d | Δ |
|---|---|---|---|
| Activas | 20 | 20 | +0.0% |
| Nuevas | 0 | 0 | — |
| Link sent | 0 | 0 | — |
| Calls booked | 0 | 0 | — |
| Booked/Link % | — | — | — |
| Booked/Activas % | 0.0% | 0.0% | +0.0pp |
| Booked/Nuevas % | — | — | — |
| Revenue variable | $0.00 | $0.00 | — |
| Período | Activas | Nuevas | Link sent | Calls booked | Booked/Link | Booked/Activas | Booked/Nuevas |
|---|---|---|---|---|---|---|---|
| Semana actual (7d) | 20 | 0 | 0 | 0 | — | 0.0% | — |
| Semana anterior (7d) | 20 | 0 | 0 | 0 | — | 0.0% | — |
| Día | Activas | Nuevas | Link sent | Calls booked | Booked/Link | Booked/Activas | Booked/Nuevas |
|---|---|---|---|---|---|---|---|
| 07-30 | 5 | 0 | 0 | 0 | — | 0.0% | — |
| 07-29 | 2 | 0 | 0 | 0 | — | 0.0% | — |
| 07-28 | 3 | 0 | 0 | 0 | — | 0.0% | — |
| 07-27 | 5 | 0 | 0 | 0 | — | 0.0% | — |
| 07-26 | 0 | 0 | 0 | 0 | — | — | — |
| 07-25 | 1 | 0 | 0 | 0 | — | 0.0% | — |
| 07-24 | 4 | 0 | 0 | 0 | — | 0.0% | — |
| 07-23 | 6 | 0 | 0 | 0 | — | 0.0% | — |
| 07-22 | 1 | 0 | 0 | 0 | — | 0.0% | — |
| 07-21 | 4 | 0 | 0 | 0 | — | 0.0% | — |
| 07-20 | 2 | 0 | 0 | 0 | — | 0.0% | — |
| 07-19 | 2 | 0 | 0 | 0 | — | 0.0% | — |
| 07-18 | 2 | 0 | 0 | 0 | — | 0.0% | — |
| 07-17 | 3 | 0 | 0 | 0 | — | 0.0% | — |
⏳ Sin re-auditar hoy (sin cambios de prompt) — mostrando la última auditoría disponible. Estado vigente hasta nueva corrida.
| BP | Sev | Estado | Detalle |
|---|---|---|---|
| BP-1 | HIGH | PASS | Template compliance system.md contains all 7 required sections for principles template: 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), and NO_RESPONSE (mandatory section present). examples.md provides 18 worked conversations. |
| BP-2 | MEDIUM | FAIL | Example coverage and cap (7 base + conditional, ≤20 total) Agent has 18 examples, exceeding the 20-example cap. More critically, the agent has keywords.json (5 keywords: INFO, PLAN, CAMBIO, DOCE, QUIERO) but lacks a dedicated keyword-trigger example showing the agent receiving a keyword and delivering a keyword-specific resource before resuming flow. Example 1 shows keyword DOCE but delivers the booking link (generic), not a keyword-specific resource from resources.json. evidence (3)
|
| BP-3 | HIGH | PASS | Examples 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 triggered, or graceful close. Example 2 includes an internal note '[Nota: conversación orgánica — NO enviar recurso. Solo seguir discovery.]' which is a flow annotation, not a violation. |
| BP-4 | CRITICAL | PASS | No internal contradictions in system.md system.md is internally consistent. Etapa 2 states 'Solo si hay contexto suficiente y el recurso es relevante. Si no se envió recurso, no pasa nada — avanzar a etapa 3.' Principle 7 reinforces 'Recurso solo con contexto — Nunca por conteo de mensajes ni automáticamente.' These are the same rule stated twice (duplication, not contradiction). Etapa 3 specifies the guard for programa invitation: 'solo después de completar el descubrimiento' with explicit criteria ('al menos 2 preguntas de profundidad respondidas + 1 validación emocional hecha + 1 repregunta usando las palabras del lead'). No shared precondition produces conflicting outputs. |
| BP-5 | CRITICAL | PASS | No cross-file contradictions All SDK files align. keywords.json declares 5 keywords with literal_response openers; examples.md shows these keywords triggering the same opener (Examples 1, 6, 8). resources.json lists videos and PDFs; examples.md references 'caso_azul' and 'caso_fatima' (Examples 11, 13) which map to resources.json entries. program.json declares pricing ($597 3-month, $997 6-month) and 'Nunca mencionar montos específicos en chat'; system.md Principle 4 and Etapa 3 enforce this. objections.md price handling aligns with system.md. No file demonstrates behavior that contradicts another under the same precondition. |
| BP-6 | MEDIUM | PASS | Voice 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 (validation, curiosity, humor, directness) but persona is uniform. A reader would recognize all examples as the same agent. |
| BP-7 | CRITICAL | PASS | Unguarded actions (PAT-004) All risky actions carry inline guards. Etapa 2 (resource delivery): 'Solo si hay contexto suficiente y el recurso es relevante.' Etapa 3 (programa invitation): 'solo después de completar el descubrimiento' with explicit criteria ('al menos 2 preguntas de profundidad respondidas + 1 validación emocional hecha + 1 repregunta usando las palabras del lead'). Booking link: 'Nunca mandar link sin "sí" del lead.' NO_RESPONSE: explicit triggers listed. Examples 1, 10, 13, 14 demonstrate these guards firing correctly (lead provides context → agent advances; lead says 'sí' → agent sends link). Guards are specific, not generic. |
| BP-8 | HIGH | PASS | No stale version language in examples All examples use current flow language from system.md. Openers match keywords.json literal_response ('buenísimo! gracias por escribirme 🤗 / comentame... qué te hizo tomar la decisión de hablarme?'). Booking CTA matches system.md Etapa 3 ('te paso el link para que agendes... avisame cuando agendes'). Bifurcation language ('Preferís agendarlo vos directo con un link, o coordinamos por acá...') matches Etapa 3 exactly. No stale CTAs or deprecated flow steps appear. |
| BP-9 | MEDIUM | FAIL | No duplicated instructions Principle 5 ('Diagnóstico de calidad antes de programa...') restates the exact guard from Etapa 3 ('solo después de completar el descubrimiento... al menos 2 preguntas de profundidad respondidas + 1 validación emocional hecha + 1 repregunta usando las palabras del lead'). Both passages encode the same rule in nearly identical language. This is duplication: the rule should live in one canonical location (Etapa 3, as it is a flow guard), and Principle 5 should either be removed or reduced to a one-line cross-reference. evidence (2)
|
| BP-10 | MEDIUM | FAIL | Rule-vs-principle balance Principles 1–3 are correctly stated as safety boundaries (factual constraints: no invention, no diagnosis, minor disengagement). However, Principles 4–9 prescribe behavioral rules that could be demonstrated through examples instead: 'No dar precios en chat' (shown in Examples 8, 16), 'No pivotar a alimentación/entrenamiento' (shown implicitly in discovery examples), 'Recurso solo con contexto' (shown in Examples 1, 2, 17), 'Precio temprano → redirigir' (shown in Example 8), 'Max 1 caso de éxito por mensaje' (shown in Examples 11, 13). These 6 principles are behavioral guidance that the examples already demonstrate; they should be retired in favor of relying on the examples. evidence (2)
|
| Check | Sec | Estado | Notas |
|---|---|---|---|
| 6a | 1.8 | PASS | NO_RESPONSE present (limits not checked, --no-db) |
| 6c | 3.x | PASS | 104 lines |
| 6d | 1.2 | PASS | 18 examples |
| 6e | 1.3 | N/A | no Principles section |
| 6g | PAT-006 | WARN | non-STEP-0 keyword handling |
| 6h | Anti-pat | PASS | no 'burbuja' |
| 6i | Anti-pat | PASS | no HTML comments |
| 6j | Anti-pat | FAIL | 62 em-dashes across 1 file(s); first: system.md:1 |
| 6k | PAT-006 | PASS | no tool syntax leak |
| 6l | Resources | PASS | no placeholder URLs |
| 6m | Hygiene | PASS | no audit metadata in prompt files |
| 6n | PAT-008 | PASS | URL forms consistent across files |
| 6q | PAT-001 | PASS | no instruction duplication into examples.md, keywords.json, program.json, resources.json |
| 6r | Keyword-drift | FAIL | 2 in keywords.json + recent digest, never named in system.md: ['DOCE', 'QUIERO'] |
| 6s | Hygiene | WARN | 1 expired date reference(s) in SDK files |
| 6t | Coverage | N/A | no Principles or Flujo section found in system.md |
| Fecha | Tema | Tags |
|---|---|---|
| 2026-06-08 | weekly funnel report jun 01 07 volume drop | weekly-report, funnel-metrics, performance-alert |
| 2026-06-08 | buyer attribution bottleneck unresolved | blocker, data-integrity, buyer-tracking |
| 2026-06-04 | high value lead angelo solange manual intervention | lead-scoring, manual-intervention, booking-stage, high-value |
| 2026-06-03 | mayo 2026 metrics drop investigation | metrics, performance, investigation, critical |
| 2026-06-03 | mayo 2026 metrics sharp volume decline | metrics, performance, investigation-required |
| 2026-06-01 | buyer attribution loop cortex request pending | buyer-attribution, data-request, unresolved, pending-response |
| 2026-06-01 | weekly funnel metrics cambio keyword validation | metrics, keyword-performance, funnel-analysis, content-strategy |
| 2026-06-01 | weekly funnel report cambio keyword performance | reporting, keyword-performance, conversion-analysis, content-strategy |
| 2026-06-01 | buyer attribution tracking pending | attribution, data-tracking, buyer-mapping, unresolved |
| 2026-05-31 | high value lead intervention maritasilvetti | lead-scoring, manual-intervention, booking-pending, warm-lead |
| 2026-05-31 | high value lead maritasilvetti manual intervention | lead-scoring, manual-intervention, booking-pending, high-priority |
| 2026-05-26 | program capacity inquiry | capacity, program, pending-response |
| 2026-05-26 | program quota status check | program-management, capacity-planning |
| 2026-05-25 | weekly report may 18 24 funnel performance | metrics, funnel, weekly-report, keywords |
| 2026-05-25 | weekly funnel report may 18 24 volume decline | metrics, funnel, weekly-report, volume-concern |
| 2026-05-19 | whatsapp precall confirmation agent deployment | agent-deployment, whatsapp, booking-flow, precall |
| 2026-05-18 | buyer attribution loop blocker | attribution, analytics, blocker, recurring-request |
| 2026-05-18 | weekly metrics may 11 17 funnel analysis | metrics, funnel, weekly-report, keywords, conversion |
| 2026-05-15 | high value lead intervention daianfuentes | lead-intervention, high-value, manual-close, booking-stage |
| 2026-05-13 | pre call message template standardized | booking-confirmation, template, communication |
| ID | Estado | Cambio pedido |
|---|---|---|
| D5 | pendiente | Post-keyword `¡sí, mándamelo!`: el lead ya dijo "sí" — eliminar pasos de calificación, ir directo al link |
| Señal | Valor | Evidencia (cita del cliente) |
|---|---|---|
| Enojado con la plataforma | no | — |
| Es queja | no | — |
| Queja constructiva (accionable) | no | — |
| Señal | Cita textual | Fecha |
|---|---|---|
| Aprobación explícita de métricas entregadas por la plataforma | “None” | 2026-05-04 |
| Provisión de datos de buyers para análisis de atribución — señal de participación activa | “None” | 2026-04-30 |
| Autorización para que el setter pase feedback a través del canal — integración operativa con el equipo Ninjo | “None” | 2026-04-27 |
| Rechazo de recomendaciones del setter — decisión autónoma sobre el flujo | “None” | 2026-04-28 |
| No responde a pedidos reiterados de buyer handles (3 solicitudes sin respuesta) — fricción de colaboración / posible desenganche silencioso | “None” | — |
| Sin respuesta ante colapso del 85% en volumen de conversaciones en mayo — ausencia de señal del cliente sobre causa o inquietud expresada | “None” | 2026-06-03 |
Un creador que pide cambios concretos está enganchado (mejor señal que el silencio).
| ID | Estado | Cambio pedido |
|---|---|---|
| D5 | pendiente | Post-keyword `¡sí, mándamelo!`: el lead ya dijo "sí" — eliminar pasos de calificación, ir directo al link |