Review — alicehongofficial

← overview

ventana últimos 7d vs 7d previos · tz America/Argentina/Buenos_Aires · último cambio Ninjo: 2026-09-28 11:00 (deploy config, hace 0d)

264

Nuevas (7d)

prior 107

Calls booked

0.0% de link

Link sent

0.9% de activas

Revenue variable

140 msgs

Resumen — período vs período

Métrica	this 7d	prior 7d	Δ
Activas	348	163	+113.5%
Nuevas	264	107	+146.7%
Link sent	3	0	—
Calls booked	0	0	—
Booked/Link %	0.0%	—	—
Booked/Activas %	0.0%	0.0%	+0.0pp
Booked/Nuevas %	0.0%	0.0%	+0.0pp
Revenue variable	$4.90	$1.93	+153.9%

Funnel — semana actual vs anterior

Período	Activas	Nuevas	Link sent	Calls booked	Booked/Link	Booked/Activas	Booked/Nuevas
Semana actual (7d)	348	264	3	0	0.0%	0.0%	0.0%
Semana anterior (7d)	163	107	0	0	—	0.0%	0.0%

Ver desglose por día

Día	Activas	Nuevas	Link sent	Booked/Link	Booked/Activas	Booked/Nuevas
09-27	57	31	0	—	0.0%	0.0%
09-26	76	66	1	0.0%	0.0%	0.0%
09-25	39	29	1	0.0%	0.0%	0.0%
09-24	51	37	1	0.0%	0.0%	0.0%
09-23	32	24	0	—	0.0%	0.0%
09-22	41	34	0	—	0.0%	0.0%
09-21	52	43	0	—	0.0%	0.0%
09-20	28	23	0	—	0.0%	0.0%
09-19	45	31	0	—	0.0%	0.0%
09-18	38	31	0	—	0.0%	0.0%
09-17	29	19	0	—	0.0%	0.0%
09-16	6	0	0	—	0.0%	—
09-15	9	3	0	—	0.0%	0.0%
09-14	8	0	0	—	0.0%	—

run 2026-07-24T14:58:26+00:00 · fuente: live DB state (daily_audit pipeline)

10/10

BP semánticos PASS

top-10 best practices

10/16

Determinísticos PASS

3 FAIL · 3 WARN

C0 · H0 · M0

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

BP	Sev	Estado	Detalle
BP-1	HIGH	PASS	Template compliance system.md contains all 7 required sections for the principles template: Identity, Voice, Calibration, Principles, Tools, NO_RESPONSE, and examples.md provides 18 worked examples. All sections match the template's specification.
BP-2	MEDIUM	PASS	Example coverage and cap (7 base scenarios + conditional, ≤20 total) examples.md contains 18 examples covering all 7 base scenarios (happy path, resource delivery, objection handling, disqualification, emotional lead, re-engagement, off-topic) plus two conditional scenarios (keyword trigger in Examples 4, 13, 18; price objection in Examples 6, 6b). Total count is 18, within the ≤20 cap.
BP-3	HIGH	PASS	Examples align with the flow Every example in examples.md follows the flow defined in system.md: greeting → discovery (max 2 questions) → resource or call → terminal state. No example skips required steps or ends mid-flow. Examples 1, 2, 3 show complete happy paths; Examples 6, 6b show price objection handling with call routing; Examples 10, 16 show NO_RESPONSE terminal states.
BP-4	CRITICAL	PASS	No internal contradictions in system.md system.md is internally consistent. Principle 3 (max 2 discovery questions, then resource) and Principle 7 (one message = one question or one next step) operate at different levels: Principle 3 governs turn count; Principle 7 governs message format. They coexist without conflict. Example 11 (marked WRONG) demonstrates the violation of Principle 3, not a contradiction between principles.
BP-5	CRITICAL	PASS	No cross-file contradictions All SDK files agree on behavior and parameters. system.md states campaign keywords take priority and send literal responses; keywords.json declares the keywords and their literal_response fields; examples.md (Examples 4, 13, 18) demonstrates keywords triggering correctly. Pricing rules in system.md (Principle 2: never say 'i dont share pricing in DMs') are consistently applied in objections.md, program.json, and examples.md (Examples 6, 6b). No file contradicts another under the same precondition.
BP-6	MEDIUM	PASS	Voice consistency across examples All examples use the same persona: warm, direct, minimal filler, emoji usage (❤️, 🔥, 💋), no bullet points, conversational tone, and consistent message length. Voice samples in system.md match the tone across all 18 examples. No jarring tonal shifts detected.
BP-7	CRITICAL	PASS	Unguarded actions (PAT-004) All risky actions in system.md have inline guards. Principle 3 guards resource sending ('max 2 discovery questions, then send resource'). Principle 4 guards campaign keyword delivery ('always check first, call lookup_keyword()... even in conversations with history. Exception: if the same link was already sent, skip the link'). Principle 5 guards link sending ('max 1 resource link per conversation. If one was already sent, nurture with text only'). Examples 1–18 demonstrate these guards firing correctly.
BP-8	HIGH	PASS	No stale version language in examples All examples use current flow language and CTAs matching system.md. Booking CTA in Examples 1, 2, 6 ('the next step is to book a call with my team...') aligns with current flow. Keyword responses in Examples 4, 13, 18 match the literal_response fields in keywords.json. No stale phrases detected.
BP-9	MEDIUM	PASS	No duplicated instructions Each behavioral instruction has a single canonical home. Principle 2 (pricing rule) is stated once in system.md and cross-referenced in objections.md and program.json. Principle 3 (max 2 discovery questions) is stated once in system.md and demonstrated in examples.md. Principle 4 (campaign keywords) is stated once in system.md and demonstrated in examples.md. No instruction is restated verbatim in two locations.
BP-10	MEDIUM	PASS	Rule-vs-principle balance Behavioral guidance (tone, flow, conversational style) is expressed via Voice section (5–6 real messages) and Principles (5 principles). Hard rules remain for factual constraints: email address (alice@alicehong.com), booking links, keyword names, and NO_RESPONSE triggers. Principles 1–5 govern behavior; Knowledge and tools govern facts. Balance is appropriate.

Checks determinísticos

Check	Sec	Estado	Notas
6a	1.8	PASS	NO_RESPONSE present (limits not checked, --no-db)
6c	3.x	PASS	69 lines
6d	1.2	PASS	22 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	FAIL	HTML comments in: ['examples.md']
6j	Anti-pat	FAIL	38 em-dashes across 2 file(s); first: system.md:1
6k	PAT-006	FAIL	tool-call syntax appears as raw text in examples.md (4 line(s))
6l	Resources	PASS	no placeholder URLs
6m	Hygiene	WARN	audit metadata (dates/slack refs) in 2 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	PASS	all 8 keyword(s) consistent across keywords.json + system.md (no DB)
6s	Hygiene	WARN	2 expired date reference(s) in SDK files
6t	Coverage	N/A	no Principles or Flujo section found in system.md

canal ninjo-alice-hong · sync 2026-09-28

108

Digests

1 en últimos 14d

Cambios pedidos / pendientes

del checklist

Threads

en el canal

2026-09-28

Última sync

del digest

Cosas nuevas (digests recientes)

Fecha	Tema	Tags
2026-09-28	keyword trigger coverage gap analysis	bot-behavior, keyword-gating, agent-configuration
2026-08-14	mcp working notes claude project setup	agent-configuration, mcp, claude-project, documentation, tooling
2026-08-03	european client geo compliance	agent-configuration, compliance, geo-restrictions
2026-08-02	eu ai labeling compliance meta enforcement	compliance, eu-regulations, meta-policy, ai-disclosure
2026-07-29	message chunking bug keyword resources	bug, agent-configuration, instagram-dm-automation, keyword-resource-delivery
2026-07-29	keyword resource delivery message chunking bug	bug, critical, message-formatting, instagram-dm, keyword-delivery
2026-07-28	ninjo workspace access path recommendation	infrastructure, access-control, setters, workflow
2026-07-27	slack channel access and ownership transfer	slack-workspace, access-control, channel-management, blocking-issue
2026-07-27	team emails shared for access	operations, access, team-emails
2026-07-25	ig lead alerts channel rename admin request	admin-request, channel-management, lead-tracking
2026-07-25	comment keyword trigger regression returning contacts	bug, instagram-dm-automation, keyword-triggers, regression
2026-07-25	api schema validation and workflow patch bugs	bug, api, agent-configuration, data-integrity
2026-06-22	ghl connection guide delivered	infrastructure, ghl, onboarding, mcp
2026-06-19	mcp setup and agent reactivation	agent-configuration, mcp-setup, infrastructure, agent-reactivation
2026-06-10	account communication and meeting request	meeting-request, partnership-alignment
2026-06-09	service model negotiation and agent value proposal	service-agreement, billing, agent-performance, partnership-terms
2026-06-08	realignment call request thursday friday	collaboration, meeting-request, strategy
2026-06-08	bot sending resources without context	bot-behavior, quality-issue, retreat-queries
2026-05-26	masterclass alert timing bug	bot-behavior, alerts, bug
2026-05-13	lead sheet update shared	lead-management, data-tracking

Cambios pedidos por el creador / pendientes

Sin pendientes en el checklist ✅

sentiment del cliente → Ninjo · scoreado ahora (2026-09-28 20:11:31) · modelo claude-sonnet-4-6 · ventana 7d

-1

Sentiment score

frustrado_pero_enganchado

Medio

Churn risk

del lado del cliente

Cambios pedidos

pendientes en checklist

Volumen cliente

agente: 264 convs

Matriz de volumen 🚨

alarma real

Cliente callado mientras el agente tiene tráfico: desenganche / churn silencioso. Abrir loop proactivo.

Sentiment del cliente

📋 El cliente está operativamente comprometido pero cargando workarounds manuales crónicos por fallas repetidas del bot; no hay citas directas de frustración verbal, pero el patrón (shutdown de emergencia, setter cubriendo gaps, llamada urgente, lista larga de bugs pendientes) indica churn risk medio-alto si los issues abiertos no se cierran. Acción recomendada: priorizar cierre de los pending proposals más críticos (keyword gap, shutoffs, DESIRE sequence), entregar un update de estado proactivo a Alice/Amir esta semana y proponer una revisión de salud quincenal para recuperar confianza antes de que la fatiga acumulada escale.

Señal	Valor	Evidencia (cita del cliente)
Enojado con la plataforma	no	—
Es queja	sí	—
Queja constructiva (accionable)	sí	—

Evidencia citada

Señal	Cita textual	Fecha
El equipo del cliente detectó y reportó por su cuenta un gap en la cobertura de keyword triggers en chats activos	“None”	2026-09-28
El cliente cubre manualmente las fallas del bot (setter escanea chats por keyword triggers perdidos), señal de fricción operativa sostenida	“None”	2026-09-28
Historial extenso de problemas recurrentes de comportamiento del bot (repetición, shutoffs, misfires, mirroring, etc.) — patrón de desgaste acumulado	“None”	—
Bot fue apagado de emergencia por errores críticos de configuración y reactivación no autorizada — evento de alta severidad	“None”	2026-01-21
El cliente propone activamente una solución técnica (LLM secundario) en lugar de escalar como queja pura — señal de engagement y ownership	“None”	2026-09-28
Se realizó una llamada urgente con Alice, Amir y Agus para tratar fixes del agente, estado del setter, KPIs y acuerdos — señal de tensión pero también de compromiso con la relación	“None”	2026-03-06

Cambios pedidos por el creador (Métrica 1)

Sin cambios pendientes en el checklist ✅