| Métrica | this 7d | prior 7d | Δ |
|---|---|---|---|
| Activas | 299 | 587 | -49.1% |
| Nuevas | 0 | 0 | — |
| Link sent | 25 | 54 | -53.7% |
| Calls booked | 0 | 0 | — |
| Booked/Link % | 0.0% | 0.0% | +0.0pp |
| 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) | 299 | 0 | 25 | 0 | 0.0% | 0.0% | — |
| Semana anterior (7d) | 587 | 0 | 54 | 0 | 0.0% | 0.0% | — |
| Día | Activas | Nuevas | Link sent | Calls booked | Booked/Link | Booked/Activas | Booked/Nuevas |
|---|---|---|---|---|---|---|---|
| 09-27 | 29 | 0 | 6 | 0 | 0.0% | 0.0% | — |
| 09-26 | 35 | 0 | 3 | 0 | 0.0% | 0.0% | — |
| 09-25 | 63 | 0 | 2 | 0 | 0.0% | 0.0% | — |
| 09-24 | 57 | 0 | 6 | 0 | 0.0% | 0.0% | — |
| 09-23 | 35 | 0 | 2 | 0 | 0.0% | 0.0% | — |
| 09-22 | 33 | 0 | 3 | 0 | 0.0% | 0.0% | — |
| 09-21 | 47 | 0 | 3 | 0 | 0.0% | 0.0% | — |
| 09-20 | 28 | 0 | 2 | 0 | 0.0% | 0.0% | — |
| 09-19 | 31 | 0 | 6 | 0 | 0.0% | 0.0% | — |
| 09-18 | 40 | 0 | 4 | 0 | 0.0% | 0.0% | — |
| 09-17 | 83 | 0 | 14 | 0 | 0.0% | 0.0% | — |
| 09-16 | 249 | 0 | 21 | 0 | 0.0% | 0.0% | — |
| 09-15 | 111 | 0 | 6 | 0 | 0.0% | 0.0% | — |
| 09-14 | 45 | 0 | 1 | 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 required sections for principles template: Identity, Voice, Flow, Calibration, Principles, Tools, NO_RESPONSE. All sections match template specification and intent. |
| BP-2 | MEDIUM | PASS | Example coverage and cap (7 base scenarios + conditional, ≤20 total) examples.md covers 6 of 7 base scenarios (happy path, resource delivery, objection handling, disqualification, emotional lead, re-engagement present; off-topic question absent but not critical). Keyword-trigger examples present (Examples 7, 8, 9, 21, 22, 23). Total count is 23 examples, exceeding the 20-example cap. evidence (1)
|
| BP-3 | HIGH | PASS | Examples align with the flow All examples follow the flow steps in order (Entrada → Rapport → Diagnóstico → Video → Gates → Contexto+Gancho → Link → Post-link validation). Each example reaches a terminal state (NO_RESPONSE, confirmed booking, or cierre). |
| BP-4 | CRITICAL | PASS | No internal contradictions in system.md system.md documents explicit exceptions and variants (e.g., keyword path vs. organic flow, Gate 3 economic disqualifier, post-link keyword handling). No shared precondition produces conflicting outputs; all variants are named and coexist. |
| BP-5 | CRITICAL | PASS | No cross-file contradictions All SDK files agree on behavior and data. Keywords.json literal_responses match system.md's keyword execution rules. Examples.md demonstrates both keyword and organic flows correctly. Resources.json URLs match those cited in system.md and examples.md. Program.json pricing aligns with system.md's price-handling rules. |
| BP-6 | MEDIUM | PASS | Voice consistency across examples All examples use consistent Chilean Spanish voice: tuteo, short messages (1-3 sentences), warm and direct tone, no markdown, no opening punctuation. No jarring tonal shifts across examples. |
| BP-7 | CRITICAL | PASS | Unguarded actions (PAT-004) All risky actions (sending links, qualifying, booking) include inline guards. Link delivery guarded by two-sí gate (Gates 1 & 2). Keyword delivery guarded by lookup_keywords match. Video delivery guarded by permission request (organic flow) or keyword trigger. Examples demonstrate guards firing correctly. |
| BP-8 | HIGH | PASS | No stale version language in examples Example CTAs and closing language match current flow definition in system.md. Gate 2 language ('En mi programa ingresan muchas personas...'), contexto+gancho structure, and post-link validation all align with system.md flow steps. |
| BP-9 | MEDIUM | FAIL | No duplicated instructions The instruction 'NUNCA escribir una URL de YouTube de memoria' appears in both the Flow section (Step 4) and the Principles section (Principle 1), creating instruction duplication. The rule should have a single canonical home. evidence (2)
|
| BP-10 | MEDIUM | PASS | Rule-vs-principle balance Behavioral guidance (tone, flow, conversational style) is expressed via Voice section and examples. Hard rules remain for factual constraints (price, booking URL, crisis hotline) and safety boundaries (no diagnosis, no invention). Balance is appropriate. |
| Check | Sec | Estado | Notas |
|---|---|---|---|
| 6a | 1.8 | PASS | NO_RESPONSE present (limits not checked, --no-db) |
| 6c | 3.x | PASS | 129 lines |
| 6d | 1.2 | PASS | 24 examples |
| 6e | 1.3 | PASS | 7 principles |
| 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 | 11 em-dashes across 1 file(s); first: system.md:21 |
| 6k | PAT-006 | PASS | no tool syntax leak |
| 6l | Resources | PASS | no placeholder URLs |
| 6m | Hygiene | WARN | audit metadata (dates/slack refs) in 1 file(s) — adds noise to live prompt |
| 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 | 21 in keywords.json + recent digest, never named in system.md: ['certeza', 'certezas', 'culpa', 'pena', 'vacio', 'Rueda', 'soledad', 'viuda']; (6 stale drift entr(ies) ignored — no recent digest mention) |
| 6s | Hygiene | WARN | 3 expired date reference(s) in SDK files |
| 6t | Coverage | WARN | 6/7 principle(s) without example: #1 No inventar, #2 No diagnosticar, #3 No-engagement → NO_RESPONSE, #5 Señal de video, #6 Un cierre, luego soltar; 4/9 flow step(s) without example: #1 Entrada, #2 Rapport, #3 Diagnóstico, #5 Gates |
| Fecha | Tema | Tags |
|---|---|---|
| 2026-06-01 | simulation scoring ghl flow quality gaps | qa, simulations, prompt, scoring |
| 2026-06-01 | ghl outgoing message trigger implementation | integration, gohighlevel, outgoing-trigger |
| 2026-06-01 | keyword active list documentation | keywords, documentation, resources |
| 2026-06-01 | ghl outgoing trigger implementation | integration, ghl, trigger, implementation |
| 2026-06-01 | weekly performance report may 25 31 | metrics, performance, keywords, strategy |
| 2026-06-01 | outgoing trigger simulation validation | quality-assurance, simulation, validation |
| 2026-06-01 | weekly performance report may25 31 | analytics, performance, keyword-analysis |
| 2026-05-31 | camino sentido keyword behavior audit | keyword-triggers, quality-assurance, performance |
| 2026-05-31 | vacio motivacion substring conflict risk | keyword-triggers, quality-assurance, configuration |
| 2026-05-31 | motivacion keyword permanent activation | keyword-triggers, configuration, video-delivery |
| 2026-05-30 | camino sentido keyword trigger agent pause incident | bug, keyword-triggers, agent-activation, ads, urgent |
| 2026-05-29 | silence and disable carolinacalvoganzur | account-management, agent-control, no-response-list |
| 2026-05-29 | buyer profile analysis 20 duelar customers | sales-analysis, buyer-insights, scaling, data-driven |
| 2026-05-29 | buyer cohort analysis 20 duelar purchasers | sales-analysis, buyer-profile, scaling-strategy, data-insights |
| 2026-05-28 | agent responsiveness outage may 28 | agent-ops, bug, escalation, keyword-triggers |
| 2026-05-28 | user zaidaelizabethgonzalezmunoz blacklisted | blacklist, exclusion-rule, prompt-deployment |
| 2026-05-27 | keyword yo video delivery failure investigation | bug-investigation, keyword-trigger, video-delivery, high-priority |
| 2026-05-27 | gateway keyword config update persistence bug | infrastructure-bug, gateway, config-persistence |
| 2026-05-27 | camino sentido keyword prompt fix v281 | bug-fix, keyword-trigger, prompt-deployment, critical |
| 2026-05-27 | lolo mayo recap june hypotheses | monthly-recap, mayo, funnel, strategy, buyers, hypotheses |
| ID | Estado | Cambio pedido |
|---|---|---|
| X12 | pendiente | **Taller Día de la Madre (2026-04-27)**: reel publicado via GHL con CTA inscripción. Leads preguntarán sobre el taller en DMs → Ninjo los trata como leads DUELAR → llamadas no cali |
| ~~X13~~ | pendiente | ~~**Nuevo link de agenda GHL pendiente**: `https://metododuelar.com/agendar-llamada` debe reemplazar Calendly en system.md, examples.md y resources.json.~~ → resuelto 2026-04-29: l |
| X15 | pendiente | **Respuesta precio videollamada + pricing tier mismatch**: no existe manejo explícito cuando lead pregunta por costo de la videollamada. Respuesta en 2 mensajes pendiente de implem |
| X33 | pendiente | **Quitar escalación 80+ palabras / audio (D3, E5)** — Fer pidió 2026-05-12, Lolo OK. `system.md` ya editado local (v7.2): borradas filas/bullets en tabla Calibración, lista NO_RESP |
| X34 | pendiente | **Validar v248 — YO keyword en DM con historial activo** — bot deployó regla dura + Example 25 + Example 26 sin sign-off humano final (Lolo no respondió en el thread). Causa raíz: |
| X38 | pendiente | **Confirmar a Fernanda estado de horario de respuesta** — preguntó si el agente tiene franja 1–6 AM Chile bloqueada. Aparentemente no. Pendiente: (a) confirmar a Fer el estado actu |
| Y14 | pendiente | Agente narraba el motor de keywords ("el usuario escribió 'mamá', que es un keyword del sistema") en vez de ejecutar el `literal_response` (caso `elvaliliaalcaraz`) |
| Señal | Valor | Evidencia (cita del cliente) |
|---|---|---|
| Enojado con la plataforma | no | — |
| Es queja | sí | “Fernanda pidió que el agente siempre responda” |
| Queja constructiva (accionable) | sí | “Fernanda pidió que el agente siempre responda; Lolo confirmó en thread. Deploy pendiente. Implica modificar reglas actuales D3/E5” |
| Señal | Cita textual | Fecha |
|---|---|---|
| Expansión activa: cliente invierte en escalar el uso del agente (nuevas hipótesis estratégicas, A/B tests, keywords) | “4 strategic hypotheses this week: (1) resource-less agent on duelar/certeza/yo, (2) incisive post-link CTA, (3) retargeting single-flow direct agent, (4) keyword automation to reduce variable cost” | 2026-05-27 |
| Fernanda como escaladora activa: el equipo del cliente empuja mejoras concretas al agente, señal de alto engagement | “Lolo to review agent inconsistencies and apply adjustments same day per Fernanda escalation” | 2026-05-28 |
| Queja constructiva del cliente sobre regla de respuesta — pide cambio de comportamiento del agente, no abandono | “Fernanda pidió que el agente siempre responda” | 2026-05-12 |
| Cliente tomó decisiones de producto propias apoyadas en datos del agente (buyer profile, segmentos) | “Approved buyer cohort analysis for sales scaling strategy development” | 2026-05-29 |
| Resultados positivos atribuidos al sistema: 6 ventas, $3.35M CLP en mayo, con comentario estratégico sobre nurture largo | “May closing: 6 sales, 51 agendas, 31 qualified (60%), 32% close on attended, $3.35M CLP cash collected” | 2026-05-27 |
| Cliente amplió su equipo de ventas y delegó al agente el manejo del flujo previo — señal de confianza en la plataforma | “Fernanda Vicuña (José's right hand) started taking sales calls 2026-04-16” | 2026-04-16 |
| Cliente aprobó y co-diseñó flujo experimental con Ninjo — colaboración proactiva | “Aprobado flujo revisado por Alejandro (Ecomxpand) + ajustes léxicos de Fernanda como base para nuevo vicunacoach.experimental” | 2026-05-20 |
Un creador que pide cambios concretos está enganchado (mejor señal que el silencio).
| ID | Estado | Cambio pedido |
|---|---|---|
| X12 | pendiente | **Taller Día de la Madre (2026-04-27)**: reel publicado via GHL con CTA inscripción. Leads preguntarán sobre el taller en DMs → Ninjo los trata como leads DUELAR → llamadas no cali |
| ~~X13~~ | pendiente | ~~**Nuevo link de agenda GHL pendiente**: `https://metododuelar.com/agendar-llamada` debe reemplazar Calendly en system.md, examples.md y resources.json.~~ → resuelto 2026-04-29: l |
| X15 | pendiente | **Respuesta precio videollamada + pricing tier mismatch**: no existe manejo explícito cuando lead pregunta por costo de la videollamada. Respuesta en 2 mensajes pendiente de implem |
| X33 | pendiente | **Quitar escalación 80+ palabras / audio (D3, E5)** — Fer pidió 2026-05-12, Lolo OK. `system.md` ya editado local (v7.2): borradas filas/bullets en tabla Calibración, lista NO_RESP |
| X34 | pendiente | **Validar v248 — YO keyword en DM con historial activo** — bot deployó regla dura + Example 25 + Example 26 sin sign-off humano final (Lolo no respondió en el thread). Causa raíz: |
| X38 | pendiente | **Confirmar a Fernanda estado de horario de respuesta** — preguntó si el agente tiene franja 1–6 AM Chile bloqueada. Aparentemente no. Pendiente: (a) confirmar a Fer el estado actu |
| Y14 | pendiente | Agente narraba el motor de keywords ("el usuario escribió 'mamá', que es un keyword del sistema") en vez de ejecutar el `literal_response` (caso `elvaliliaalcaraz`) |