opencode-skills-collection 4.0.52 → 4.0.54

package/bundled-skills/analytics-product/SKILL.md

@@ -27,7 +27,8 @@ Analytics de produto — PostHog, Mixpanel, eventos, funnels, cohorts, retencao,
 
 ## When to Use This Skill
 
-- When you need specialized assistance with this domain
+- Use para definir um evento de ativacao, investigar queda de funil ou calcular retencao com denominador e janela explicitos.
+- Antes de instrumentar, registre a decisao de produto, a fonte de dados, o consentimento aplicavel, o fuso horario e a unidade de analise.
 
 ## Do Not Use This Skill When
 
@@ -50,7 +51,7 @@ Errado:    signup, click, conversion
 
 ---
 
-## Eventos Essenciais Da Auri
+## Exemplo ilustrativo: eventos de um assistente
 
 ```python
 AURI_EVENTS = {
@@ -113,7 +114,7 @@ track("user_123", "conversation_started", {
 
 ---
 
-## Funil De Ativacao Auri
+## Funil ilustrativo de ativacao (numeros hipoteticos)
 
 ```
 Visita landing page          (100%)
@@ -139,7 +140,8 @@ Para cada drop-off > benchmark:
 2. Entender: por que? (session recordings, surveys)
 3. Hipotese: qual mudanca poderia melhorar?
 4. Testar: A/B test com amostra estatisticamente significante
-5. Medir: 2 semanas minimo, p-value < 0.05
+5. Medir: janela e amostra predefinidas, efeito com intervalo, qualidade e guardrails
+   Nao encerrar cedo por um p-value favoravel; investigar SRM e perdas de tracking
 6. Aprender: mesmo se falhar, entende-se o usuario melhor
 ```
 
@@ -172,9 +174,11 @@ def calculate_cohort_retention(events_df):
     return retention
 ```
 
-## Benchmarks De Retencao (Assistentes De Voz)
+## Faixas ilustrativas de retencao (nao sao benchmarks de mercado)
 
-| Semana | Pessimo | Ok | Bom | Excelente |
+Estes numeros nao possuem fonte ou validacao externa. Use apenas como exemplo de formato; substitua por baseline observado de cohorts comparaveis e maturas.
+
+| Semana | Faixa A | Faixa B | Faixa C | Faixa D |
 |--------|---------|-----|-----|-----------|
 | W1 | <20% | 20-35% | 35-50% | >50% |
 | W4 | <10% | 10-20% | 20-30% | >30% |
@@ -182,12 +186,12 @@ def calculate_cohort_retention(events_df):
 
 ---
 
-## Definindo A North Star Da Auri
+## Hipotese ilustrativa de North Star
 
 ```
 Framework:
 1. O que cria valor real para o usuario? -> Conversas que geram insight/acao
-2. O que prediz crescimento de longo prazo? -> Usuarios com 3+ conv/semana
+2. Hipotese a validar: usuarios com 3+ conversas/semana recebem valor recorrente
 3. Como medir? -> "Weekly Active Conversationalists" (WAC)
 
 North Star: WAC (Weekly Active Conversationalists)
@@ -199,17 +203,21 @@ Meta Ano 2: 100.000 WAC
 
 ## Dashboard North Star
 
+Sketch: adapte `db.query` e `calculate_wow_growth` ao projeto. Use limites de janela explicitos e o mesmo fuso; conte usuarios qualificados no resultado agregado, nao uma linha por usuario.
+
 ```python
-def calculate_north_star(db):
+def calculate_north_star(db, window_start, window_end):
     wac = db.query("""
-        SELECT COUNT(DISTINCT user_id) as wac
-        FROM conversations
-        WHERE
-            created_at >= NOW() - INTERVAL '7 days'
-            AND duration_seconds >= 120
-        GROUP BY user_id
-        HAVING COUNT(*) >= 3
-    """).scalar()
+        SELECT COUNT(*) as wac
+        FROM (
+            SELECT user_id
+            FROM conversations
+            WHERE created_at >= :window_start AND created_at < :window_end
+              AND duration_seconds >= 120
+            GROUP BY user_id
+            HAVING COUNT(*) >= 3
+        ) AS qualifying_users
+    """, {"window_start": window_start, "window_end": window_end}).scalar()
 
     return {
         "wac": wac,
@@ -223,9 +231,12 @@ def calculate_north_star(db):
 
 ## Feature Flags Com Posthog
 
+Use a API da versao instalada. O SDK atual oferece `evaluate_flags`; em versoes antigas, a ordem de `feature_enabled` era `(feature, user_id)`. Em erro ou ausencia de valor, preserve o fluxo de controle seguro. Veja a [documentacao Python oficial](https://posthog.com/docs/libraries/python). Nao envie eventos/identificacao antes da autorizacao e das regras de consentimento do projeto.
+
 ```python
 def is_feature_enabled(user_id: str, feature: str) -> bool:
-    return posthog.feature_enabled(feature, user_id)
+    flags = posthog.evaluate_flags(user_id)
+    return flags.is_enabled(feature) is True
 
 if is_feature_enabled(user_id, "new-onboarding-v2"):
     show_new_onboarding()
@@ -237,7 +248,6 @@ else:
 
 ```python
 from scipy import stats
-import numpy as np
 
 def ab_test_significance(
     control_conversions: int,
@@ -246,30 +256,43 @@ def ab_test_significance(
     variant_visitors: int,
     confidence: float = 0.95
 ) -> dict:
+    counts = (control_conversions, control_visitors, variant_conversions, variant_visitors)
+    if any(type(value) is not int or value < 0 for value in counts):
+        raise ValueError("Contagens devem ser inteiros nao negativos")
+    if not (0 < control_visitors and 0 < variant_visitors
+            and control_conversions <= control_visitors
+            and variant_conversions <= variant_visitors and 0 < confidence < 1):
+        raise ValueError("Denominadores, conversoes ou confianca invalidos")
     control_rate = control_conversions / control_visitors
     variant_rate = variant_conversions / variant_visitors
-    lift = (variant_rate - control_rate) / control_rate * 100
+    lift = (variant_rate - control_rate) / control_rate * 100 if control_rate else None
 
-    _, p_value = stats.chi2_contingency([
+    table = [
         [control_conversions, control_visitors - control_conversions],
         [variant_conversions, variant_visitors - variant_conversions]
-    ])[:2]
+    ]
+    if any(sum(row) == 0 for row in zip(*table)):
+        return {"status": "insufficient-variation", "recommendation": "No automatic decision"}
+    _, p_value, _, expected = stats.chi2_contingency(table)
+    if (expected < 5).any():
+        return {"status": "sparse-counts", "recommendation": "Use a pre-specified exact method"}
 
     significant = p_value < (1 - confidence)
 
     return {
         "control_rate": f"{control_rate*100:.2f}%",
         "variant_rate": f"{variant_rate*100:.2f}%",
-        "lift": f"{lift:+.1f}%",
+        "lift": f"{lift:+.1f}%" if lift is not None else None,
         "p_value": round(p_value, 4),
         "significant": significant,
-        "recommendation": "Deploy variant" if significant and lift > 0 else "Keep control"
+        "absolute_difference_pp": (variant_rate - control_rate) * 100,
+        "recommendation": "Review pre-specified effect, uncertainty and guardrails; no automatic deploy"
     }
 ```
 
 ---
 
-## 6. Comandos
+## 6. Sugestoes de prompts (nao instalam comandos no cliente)
 
 | Comando | Acao |
 |---------|------|
@@ -281,26 +304,16 @@ def ab_test_significance(
 | `/dashboard-setup` | Cria dashboard de produto |
 | `/okr-template` | Template de OKRs para produto |
 
-## Best Practices
-
-- Provide clear, specific context about your project and requirements
-- Review all suggestions before applying them to production code
-- Combine with other complementary skills for comprehensive analysis
+## Exemplo verificavel
 
-## Common Pitfalls
+Entrada sintetica: em uma janela fechada, usuario A tem tres conversas de 120 segundos, B tem duas e C tem quatro de 60 segundos. O resultado WAC esperado e **1**, nao varias linhas com valor 1. Em retencao, reporte tamanho da cohort e idade observavel; uma semana ainda nao encerrada nao representa zero retencao.
 
-- Using this skill for tasks outside its domain expertise
-- Applying recommendations without understanding your specific context
-- Not providing enough project context for accurate analysis
-
-## Related Skills
-
-- `growth-engine` - Complementary skill for enhanced analysis
-- `monetization` - Complementary skill for enhanced analysis
-- `product-design` - Complementary skill for enhanced analysis
-- `product-inventor` - Complementary skill for enhanced analysis
+Para um experimento, registre unidade de randomizacao, metrica primaria, janela, efeito minimo, regra de parada e guardrails antes de calcular o teste. O exemplo de significancia rejeita denominadores invalidos e contagens esparsas; ele nao e um mecanismo de decisao de rollout.
 
 ## Limitations
-- Use this skill only when the task clearly matches the scope described above.
-- Do not treat the output as a substitute for environment-specific validation, testing, or expert review.
-- Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.
+
+- As metas, faixas e eventos de assistente acima sao hipoteticos; nao provam benchmarks ou comportamento dos usuarios.
+- O trecho de cohort assume timestamps ja normalizados e dados completos; semanas imaturas precisam ser mascaradas e cohorts sem usuarios nao devem dividir por zero.
+- Um p-value isolado nao mede valor do produto, elimina vieses ou substitui intervalos e desenho experimental.
+- SDKs podem enviar dados para servicos externos. Minimize propriedades, evite texto de conversas e valide consentimento, residencia e retencao antes de ativar tracking.
+- Os exemplos de banco e interface dependem de adaptadores do projeto; nao representam uma aplicacao pronta.

package/bundled-skills/analytics-tracking/SKILL.md

@@ -17,9 +17,9 @@ You do **not** treat GA4 numbers as truth unless validated.
 
 ---
 
-## Phase 0: Measurement Readiness & Signal Quality Index (Required)
+## Phase 0: Measurement Evidence and Optional Review Rubric
 
-Before adding or changing tracking, calculate the **Measurement Readiness & Signal Quality Index**.
+Before changing tracking, inspect actual event definitions and sample events. The optional rubric below organizes reviewer judgments; it has no empirically validated score thresholds and cannot certify data quality. Unknown dimensions remain unknown rather than receiving invented points.
 
 ### Purpose
 
@@ -27,7 +27,7 @@ This index answers:
 
 > **Can this analytics setup produce reliable, decision-grade insights?**
 
-It prevents:
+Use it to identify possible:
 
 * event sprawl
 * vanity tracking
@@ -109,22 +109,22 @@ This is a **diagnostic score**, not a performance KPI.
 
 ---
 
-### Readiness Bands (Required)
+### Illustrative planning bands (not validation gates)
 
 | Score  | Verdict               | Interpretation                    |
 | ------ | --------------------- | --------------------------------- |
-| 85–100 | **Measurement-Ready** | Safe to optimize and experiment   |
+| 85–100 | **Measurement-Ready** | Review whether observed evidence supports the intended decision   |
 | 70–84  | **Usable with Gaps**  | Fix issues before major decisions |
 | 55–69  | **Unreliable**        | Data cannot be trusted yet        |
 | <55    | **Broken**            | Do not act on this data           |
 
-If verdict is **Broken**, stop and recommend remediation first.
+Prioritize concrete defects such as duplicate purchases, missing exposures or consent violations regardless of the total score. A high score must never override a failed reconciliation.
 
 ---
 
 ## Phase 1: Context & Decision Definition
 
-(Proceed only after scoring)
+(Start from the product decision and available evidence)
 
 ### 1. Business Context
 
@@ -354,7 +354,7 @@ Analytics that violate trust undermine optimization.
 
 ### Measurement Strategy Summary
 
-* Measurement Readiness Index score + verdict
+* Observed reconciliation results, unknowns and optional subjective rubric
 * Key risks and gaps
 * Recommended remediation order
 
@@ -402,9 +402,18 @@ Analytics that violate trust undermine optimization.
 ---
 
 ## When to Use
-This skill is applicable to execute the workflow or actions described in the overview.
+
+Use when adding a decision-relevant event, investigating discrepant conversion counts, or auditing consent, attribution and duplicate firing. Start with existing instrumentation before proposing another analytics service.
+
+## Worked example
+
+Input: the UI fires `purchase_completed` on both redirect and reload. Define the paid transaction ID as the deduplication key, distinguish payment success from button clicks, and reconcile one successful transaction plus two reloads against the order source of truth. Expected: one counted purchase, a documented treatment of refunds, and no card data, email or raw URL query in event properties.
+
+Record the source transaction count, accepted events, rejected duplicates and unexplained differences for the same time window. Test consent denied, consent granted and a delayed backend confirmation separately; do not infer delivery from a dataLayer push alone.
 
 ## Limitations
-- Use this skill only when the task clearly matches the scope described above.
-- Do not treat the output as a substitute for environment-specific validation, testing, or expert review.
-- Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.
+
+- Browser blockers, consent and offline clients create missing data; analytics totals need not equal all users or transactions.
+- Attribution models describe assigned credit, not causal impact.
+- Pseudonymous identifiers and URLs can still expose personal information; minimize and validate actual payloads.
+- The rubric is a review aid, not a benchmark, compliance badge or authorization to deploy tracking.

package/bundled-skills/analyze-project/SKILL.md

@@ -3,6 +3,7 @@ name: analyze-project
 description: Forensic root cause analyzer for Antigravity sessions. Classifies scope deltas, rework patterns, root causes, hotspots, and auto-improves prompts/health.
 risk: critical
 source: community
+date_added: "2026-09-04"
 version: "1.0"
 tags: [analysis, diagnostics, meta, root-cause, project-health, session-review]
 ---