claude-code-arcane 1.3.0 → 2.0.0

package/skills/commerce-analytics/SKILL.md

@@ -0,0 +1,85 @@
+---
+name: commerce-analytics
+description: "Define and audit e-commerce analytics — conversion rate, AOV, CLV, cohort/funnel metrics, and the standard GA4 ecommerce event spec (view_item, add_to_cart, begin_checkout, purchase). Includes a dataLayer event validator. Triggers: defining commerce KPIs, instrumenting GA4 ecommerce events, auditing tracking."
+argument-hint: "[kpis | events | audit]"
+category: "ecommerce"
+user-invocable: true
+allowed-tools: Read, Glob, Grep, Bash
+---
+
+# Commerce Analytics
+
+You cannot optimize what you mis-measure. Define the KPIs that actually move a store and instrument the standard ecommerce event spec correctly — platform-agnostic, with a validator for GA4 dataLayer events.
+
+## Before Starting
+
+Gather: current tracking (GA4/GTM/other), business model (one-time vs subscription), key conversion actions, and goal (define KPIs, design/instrument events, or audit existing tracking).
+
+## How This Skill Works
+
+| Mode | Trigger | Output |
+|------|---------|--------|
+| **KPIs** | `kpis` | KPI framework + dashboard spec |
+| **Events** | `events` | Ecommerce event taxonomy + dataLayer spec |
+| **Audit** | `audit` | Tracking gap + data-quality review |
+
+---
+
+## Phase 1 — Core KPIs
+
+| Metric | Formula | Watch for |
+|--------|---------|-----------|
+| **Conversion rate** | orders / sessions | Segment by device/channel |
+| **AOV** | revenue / orders | Moves with promos/bundles |
+| **CLV** | avg order value × frequency × lifespan | The number that justifies CAC |
+| **Cart abandonment** | 1 − (checkouts / carts) | Diagnoses checkout friction |
+| **Repeat rate** | repeat customers / total | Retention health |
+| **Sell-through** | units sold / units received | Feeds `inventory-stock` forecasts |
+
+**Rule:** Pair every acquisition metric with a margin metric — revenue growth on negative contribution margin is a trap.
+
+---
+
+## Phase 2 — The Ecommerce Event Spec
+
+The standard GA4 ecommerce funnel events, each carrying `currency` + an `items[]` array:
+
+`view_item_list → view_item → add_to_cart → begin_checkout → add_shipping_info → add_payment_info → purchase` (plus `remove_from_cart`, `refund`).
+
+Each `items[]` entry: `item_id`, `item_name`, `price`, `quantity` (+ optional `item_brand`, `item_category`, `item_variant`). `purchase` also needs `transaction_id` and `value`.
+
+Validate a captured dataLayer against this spec:
+
+```bash
+python skills/commerce-analytics/scripts/ga4_ecommerce_validate.py            # demo events
+python skills/commerce-analytics/scripts/ga4_ecommerce_validate.py events.json
+```
+
+It reports PASS/FAIL per event with the missing required fields. For the full taxonomy and parameter reference, see [references/ga4-ecommerce-events.md](references/ga4-ecommerce-events.md).
+
+---
+
+## Phase 3 — Audit & Data Quality
+
+Common faults: missing `transaction_id` (duplicate purchases), `value` excluding/including tax inconsistently, `items[]` absent on purchase, double-firing (GTM + hardcoded). Cross-check store revenue vs GA4 revenue — variance > 5% means a tracking bug.
+
+---
+
+## Output Artifacts
+
+| Request | Output |
+|---------|--------|
+| KPIs | KPI framework + dashboard layout |
+| Events | Event taxonomy + dataLayer spec |
+| Audit | Gap analysis + data-quality scorecard |
+
+Commerce tracking is **COMPLIANT** when the full funnel fires with valid `items[]` and `currency`, `purchase` carries a unique `transaction_id`, revenue reconciles within 5%, and KPIs pair growth with margin.
+
+---
+
+## Related Skills
+
+- **analytics-tracking**: general GA4/GTM setup this builds on for SaaS-style funnels.
+- **inventory-stock**: sell-through metrics feed restock forecasting.
+- **commerce-data-model**: item taxonomy maps to `items[]` parameters.
+- **ab-test-setup** / **campaign-analytics**: experimentation and campaign measurement.

package/skills/commerce-analytics/references/ga4-ecommerce-events.md

@@ -0,0 +1,63 @@
+# GA4 Ecommerce Event Taxonomy & Parameter Reference
+
+The standard funnel. Fire these with a consistent `items[]` shape so reports, funnels, and audiences work. The `ga4_ecommerce_validate.py` script checks a captured dataLayer against this spec.
+
+## The funnel
+```
+view_item_list → view_item → add_to_cart → view_cart → begin_checkout
+  → add_shipping_info → add_payment_info → purchase
+```
+Plus: `select_item`, `remove_from_cart`, `refund`.
+
+## Required parameters per event
+| Event | Required (beyond items[]) |
+|-------|---------------------------|
+| view_item | currency |
+| add_to_cart / remove_from_cart / view_cart | currency |
+| begin_checkout | currency |
+| add_shipping_info | currency |
+| add_payment_info | currency |
+| purchase | currency, value, transaction_id |
+| refund | currency, transaction_id |
+| view_item_list / select_item | (item_list_id / item_list_name recommended) |
+
+## items[] object
+| Field | Req? | Notes |
+|-------|------|-------|
+| item_id | yes | Map to SKU |
+| item_name | yes | Product/variant name |
+| price | recommended | Unit price (number) |
+| quantity | recommended | Defaults to 1 |
+| item_brand | optional | |
+| item_category | optional | Up to item_category5 |
+| item_variant | optional | e.g. "Red / M" |
+| discount | optional | Per-unit discount |
+| index | optional | Position in list |
+
+## Worked example — purchase
+```json
+{
+  "event": "purchase",
+  "ecommerce": {
+    "currency": "USD",
+    "value": 42.00,
+    "transaction_id": "T-1001",
+    "tax": 3.50,
+    "shipping": 5.00,
+    "items": [
+      { "item_id": "ACME-TEE-RED-M", "item_name": "Acme Cotton Tee",
+        "item_variant": "Red / M", "price": 21.00, "quantity": 2 }
+    ]
+  }
+}
+```
+
+## Common faults (audit checklist)
+- [ ] `purchase` missing `transaction_id` → duplicate purchases on refresh
+- [ ] `value` inconsistent about tax/shipping inclusion (pick one, document it)
+- [ ] `items[]` absent or empty on funnel events
+- [ ] Double-firing (GTM tag + hardcoded gtag)
+- [ ] Currency missing → revenue not attributed
+- [ ] Store revenue vs GA4 revenue variance > 5% → tracking bug
+
+Verdict: **COMPLIANT** when the full funnel fires with valid `items[]` + `currency`, `purchase` has a unique `transaction_id`, and revenue reconciles within 5%.

package/skills/commerce-analytics/scripts/ga4_ecommerce_validate.py

@@ -0,0 +1,134 @@
+"""
+ga4_ecommerce_validate.py — Validate a captured dataLayer against the GA4
+standard ecommerce event spec.
+
+Checks each event for: a known ecommerce event name, required top-level
+parameters (currency/value/transaction_id where applicable), a non-empty
+items[] array, and required per-item fields. No external dependencies.
+
+Usage:
+    python ga4_ecommerce_validate.py                 # built-in demo events
+    python ga4_ecommerce_validate.py events.json     # your captured dataLayer
+
+Input JSON: either a single event object or an array of event objects, each:
+    {"event": "purchase", "ecommerce": {"currency": "USD", "value": 42.0,
+        "transaction_id": "T123", "items": [{"item_id": "SKU1",
+        "item_name": "Tee", "price": 21.0, "quantity": 2}, ...]}}
+
+Some setups put params at the top level instead of under "ecommerce"; both
+shapes are accepted.
+"""
+
+import json
+import sys
+
+
+# event name -> required top-level (ecommerce) params beyond items[]
+EVENT_REQS = {
+    "view_item_list": [],
+    "view_item": ["currency"],
+    "select_item": [],
+    "add_to_cart": ["currency"],
+    "remove_from_cart": ["currency"],
+    "view_cart": ["currency"],
+    "begin_checkout": ["currency"],
+    "add_shipping_info": ["currency"],
+    "add_payment_info": ["currency"],
+    "purchase": ["currency", "value", "transaction_id"],
+    "refund": ["currency", "transaction_id"],
+}
+EVENTS_REQUIRING_ITEMS = set(EVENT_REQS) - {"refund"}
+ITEM_REQUIRED = ["item_id", "item_name"]  # GA4 needs id or name; we require both for quality
+ITEM_RECOMMENDED = ["price", "quantity"]
+
+
+DEMO_EVENTS = [
+    {"event": "view_item", "ecommerce": {"currency": "USD",
+        "items": [{"item_id": "ACME-TEE", "item_name": "Acme Tee", "price": 21.0}]}},
+    {"event": "add_to_cart", "ecommerce": {"currency": "USD",
+        "items": [{"item_id": "ACME-TEE", "item_name": "Acme Tee", "price": 21.0, "quantity": 2}]}},
+    {"event": "purchase", "ecommerce": {"currency": "USD", "value": 42.0, "transaction_id": "T-1001",
+        "items": [{"item_id": "ACME-TEE", "item_name": "Acme Tee", "price": 21.0, "quantity": 2}]}},
+    # intentionally broken: missing transaction_id + value, item missing name
+    {"event": "purchase", "ecommerce": {"currency": "USD",
+        "items": [{"item_id": "ACME-MUG", "price": 9.0}]}},
+    {"event": "add_to_cart", "ecommerce": {  # missing currency + empty items
+        "items": []}},
+]
+
+
+def get_ecom(event):
+    """Return the param dict — under 'ecommerce' or at the top level."""
+    if isinstance(event.get("ecommerce"), dict):
+        return event["ecommerce"]
+    return {k: v for k, v in event.items() if k != "event"}
+
+
+def validate_event(event, idx):
+    errors = []
+    name = event.get("event")
+    if not name:
+        return [f"event[{idx}]: missing 'event' name"], None
+    if name not in EVENT_REQS:
+        return [f"event[{idx}] '{name}': not a recognized GA4 ecommerce event"], name
+
+    ecom = get_ecom(event)
+
+    for param in EVENT_REQS[name]:
+        if param not in ecom or ecom[param] in (None, ""):
+            errors.append(f"missing required param '{param}'")
+
+    if name in EVENTS_REQUIRING_ITEMS:
+        items = ecom.get("items")
+        if not isinstance(items, list) or not items:
+            errors.append("missing or empty 'items[]'")
+        else:
+            for i, it in enumerate(items):
+                missing = [f for f in ITEM_REQUIRED if not it.get(f)]
+                if missing:
+                    errors.append(f"items[{i}]: missing {', '.join(missing)}")
+                rec_missing = [f for f in ITEM_RECOMMENDED if it.get(f) in (None, "")]
+                if rec_missing:
+                    errors.append(f"items[{i}]: (recommended) missing {', '.join(rec_missing)}")
+    return errors, name
+
+
+def main():
+    if len(sys.argv) > 1:
+        path = sys.argv[1]
+        try:
+            with open(path, encoding="utf-8") as f:
+                data = json.load(f)
+        except (OSError, json.JSONDecodeError) as e:
+            print(f"[ERROR] cannot read {path}: {e}")
+            sys.exit(1)
+    else:
+        print("[INFO] No JSON given - using built-in demo events.\n")
+        data = DEMO_EVENTS
+
+    events = data if isinstance(data, list) else [data]
+
+    total_fail = 0
+    for idx, event in enumerate(events):
+        errors, name = validate_event(event, idx)
+        hard = [e for e in errors if "(recommended)" not in e]
+        if hard:
+            total_fail += 1
+            print(f"FAIL  [{idx}] {name or '?'}")
+            for e in errors:
+                print(f"        - {e}")
+        else:
+            warn = [e for e in errors if "(recommended)" in e]
+            print(f"PASS  [{idx}] {name}")
+            for w in warn:
+                print(f"        ~ {w}")
+
+    print()
+    passed = len(events) - total_fail
+    verdict = "COMPLIANT" if total_fail == 0 else "NON-COMPLIANT"
+    print(f"{verdict}: {passed}/{len(events)} events passed.")
+    sys.exit(0 if total_fail == 0 else 1)
+
+
+if __name__ == "__main__":
+    main()

package/skills/commerce-data-model/SKILL.md

@@ -0,0 +1,84 @@
+---
+name: commerce-data-model
+description: "Design and audit the core e-commerce domain model — products, variants, SKUs, options, collections, attributes, taxonomy, faceted search and pagination. Language- and platform-agnostic. Triggers: modeling a catalog, reviewing an existing product schema, designing filtering/search."
+argument-hint: "[model | audit | taxonomy]"
+category: "ecommerce"
+user-invocable: true
+allowed-tools: Read, Glob, Grep, Write, Edit
+---
+
+# Commerce Data Model
+
+Design the backbone of any storefront: a product catalog model that survives variants, bundles, localization, and faceted search — independent of language, framework, or database.
+
+> **Note:** This skill may create or modify files in your project (a model spec or schema doc). It will ask before writing.
+
+## Before Starting
+
+Gather: business context (physical/digital/services, single vs multi-vendor, B2C/B2B), scale (SKU count, variant explosion, locales/currencies), and goal (model from scratch, audit an existing schema, or design taxonomy + filtering).
+
+## How This Skill Works
+
+| Mode | Trigger | Output |
+|------|---------|--------|
+| **Model** | `model` | Entity model spec: Product, Variant, SKU, Option, Collection, Attribute |
+| **Audit** | `audit` | Gap analysis of an existing schema against commerce best practices |
+| **Taxonomy** | `taxonomy` | Category tree, attribute schema, faceted-search and pagination design |
+
+---
+
+## Phase 1 — Core Entities
+
+The canonical separation that prevents 90% of catalog pain:
+
+| Entity | Is | Is NOT | Key fields |
+|--------|----|--------|-----------|
+| **Product** | The marketing/SEO unit ("Acme Tee") | A buyable thing | title, description, slug, status, brand |
+| **Variant** | A buyable combination ("Tee / Red / M") | A category | option values, price, weight, barcode |
+| **SKU** | The stock-keeping identifier | A variant's price | code, variant_id |
+| **Option** | An axis of variation (Color, Size) | A value | name, position |
+| **Collection** | A merchandising grouping | A taxonomy node | title, rules (manual/automatic) |
+
+**Rule:** Price, inventory, and barcode live on the **Variant**, never the Product. A product with no options still has exactly one variant.
+
+For the full field-level spec, identifiers (GTIN/MPN), bundles/kits, and digital vs physical modeling, see [references/data-model-spec.md](references/data-model-spec.md).
+
+---
+
+## Phase 2 — Attributes & Taxonomy
+
+- **Taxonomy** (category tree): hierarchical, navigational, one primary path per product. Stable URLs.
+- **Collections**: many-to-many merchandising overlays (manual or rule-based). A product lives in one category branch but many collections.
+- **Attributes**: typed key-values (`material: cotton`, `sleeve: short`). Drive facets. Define a controlled vocabulary — free-text attributes break filtering.
+
+**Attribute typing:** `text | enum | number | boolean | measurement`. Enums must reference a value list. Measurements carry a unit.
+
+---
+
+## Phase 3 — Faceted Search & Pagination
+
+- Facets come from `enum`/`boolean`/`measurement` attributes, not free text.
+- Multi-select within a facet = OR; across facets = AND.
+- Always show counts; never show zero-result facet values.
+- **Pagination:** prefer cursor/keyset over offset for deep catalogs (offset degrades and double-counts under writes). Cap page size.
+
+---
+
+## Output Artifacts
+
+| Request | Output |
+|---------|--------|
+| Model from scratch | Entity spec + ER sketch + identifier strategy |
+| Audit schema | Verdict per entity: COMPLIANT / CONCERNS, with fixes |
+| Design taxonomy | Category tree + attribute schema + facet plan |
+
+A schema is **COMPLIANT** when price/inventory sit on variants, every product has ≥1 variant, attributes are typed against vocabularies, and URLs are stable.
+
+---
+
+## Related Skills
+
+- **storefront-architecture**: consumes this model to design PLP/PDP pages. Next step after modeling.
+- **merchant-feed**: maps these entities to Google/Meta feed attributes.
+- **inventory-stock**: owns stock levels that hang off the SKU/variant.
+- **order-lifecycle**: references variants as order line items.

package/skills/commerce-data-model/references/data-model-spec.md

@@ -0,0 +1,79 @@
+# Commerce Data Model — Field-Level Spec & Worked Example
+
+A language- and database-agnostic reference for the core catalog entities. Use it as the template for a model spec.
+
+## Entities & Fields
+
+### Product (marketing/SEO unit)
+| Field | Type | Notes |
+|-------|------|-------|
+| id | id | Internal key |
+| title | text | Display name |
+| slug | text | Stable URL segment, unique |
+| description | rich text | Long copy |
+| brand | text | Brand/manufacturer |
+| status | enum | draft / active / archived |
+| options | Option[] | Axes of variation (0..n) |
+| variants | Variant[] | ≥1 always |
+| collections | Collection[] | Many-to-many |
+| primary_category | Category | One navigational path |
+| attributes | Attribute[] | Typed key-values |
+
+### Variant (buyable unit)
+| Field | Type | Notes |
+|-------|------|-------|
+| id | id | |
+| sku | text | Stock-keeping code, unique |
+| option_values | map | e.g. {Color: Red, Size: M} |
+| price | integer (minor units) | Never a float |
+| compare_at_price | integer | For showing discounts |
+| barcode | text | GTIN/UPC/EAN |
+| weight / dimensions | number + unit | For shipping/carrier rates |
+| requires_shipping | bool | False for digital |
+
+### Option / Option Value
+`Option { name, position }` → `OptionValue { option_id, value, position }`. Example: Option "Size" → values S, M, L.
+
+### Collection (merchandising)
+`Collection { title, slug, type: manual|automatic, rules? }`. Automatic collections select by attribute rules (e.g. `tag = summer`).
+
+### Attribute (faceting)
+`Attribute { key, type: text|enum|number|boolean|measurement, value, unit? }`. Enum attributes reference a controlled value list.
+
+## Identifiers
+- **SKU** — your internal stock identifier (required, unique).
+- **GTIN** (UPC/EAN/ISBN) — global trade identifier; required for marketplaces/feeds where one exists.
+- **MPN** — manufacturer part number; fallback when no GTIN.
+
+## Bundles / Kits
+Model as a Product whose variant references component variants with quantities. Decide inventory rule: track by components (decrement each) vs by kit (own stock).
+
+## Worked Example — "Acme Tee"
+
+```yaml
+product:
+  title: "Acme Cotton Tee"
+  slug: "acme-cotton-tee"
+  brand: "Acme"
+  status: active
+  options:
+    - { name: Color, values: [Red, Blue] }
+    - { name: Size,  values: [S, M, L] }
+  attributes:
+    - { key: material, type: enum, value: cotton }
+    - { key: sleeve,   type: enum, value: short }
+  variants:
+    - sku: ACME-TEE-RED-M
+      option_values: { Color: Red, Size: M }
+      price: 2100            # 21.00 in minor units
+      barcode: "0123456789012"
+      weight: { value: 180, unit: g }
+```
+
+## Checklist (verdict: COMPLIANT when all true)
+- [ ] Price + inventory + barcode live on the variant, not the product
+- [ ] Every product has ≥1 variant (even with no options)
+- [ ] Attributes are typed and enums reference a vocabulary
+- [ ] Primary category is single; collections are many-to-many
+- [ ] Money stored as integer minor units
+- [ ] Slugs/URLs stable and unique

package/skills/cover-letter/SKILL.md

@@ -7,9 +7,9 @@ user-invocable: true
 allowed-tools: Read, Glob, Grep, Write, Edit, WebFetch
 ---
 
-# Cover Letter — Mensaje de aplicación de élite
+# Cover Letter - Mensaje de aplicación de élite
 
-Escribís cover letters y mensajes de aplicación **concisos, seguros y custom** — que suenan a una persona que entendió el rol y la empresa, no a una plantilla. El objetivo: que el lector quiera leer el CV.
+Escribís cover letters y mensajes de aplicación **concisos, seguros y custom** - que suenan a una persona que entendió el rol y la empresa, no a una plantilla. El objetivo: que el lector quiera leer el CV.
 
 ## Inputs
 
@@ -20,19 +20,19 @@ Escribís cover letters y mensajes de aplicación **concisos, seguros y custom**
 
 ## Principios
 
-- **Una página máximo; idealmente 150–250 palabras** para mensajes, 250–350 para carta formal.
+- **Una página máximo; idealmente 150-250 palabras** para mensajes, 250-350 para carta formal.
 - **El primer renglón gana o pierde.** Nada de "I am writing to apply for…". Abrí con un gancho específico: un logro relevante, un punto de conexión con la empresa, o por qué este rol te mueve.
-- **Mostrá fit, no historia.** 1–2 logros que mapean directo a lo que piden, con número.
+- **Mostrá fit, no historia.** 1-2 logros que mapean directo a lo que piden, con número.
 - **Específico de la empresa.** Una frase que demuestre que investigaste (producto, valor, problema que resuelven). Si la podés pegar en otra aplicación, no sirve.
 - **Confiado, no arrogante; cálido, no servil.** Ni "would be honored", ni "I'm the best candidate".
 - **CTA simple al final.** Disponibilidad / ganas de charlar, sin rogar.
 
 ## Estructura (carta formal)
 
-1. **Hook** — gancho específico (logro o conexión real).
-2. **Fit** — 1–2 logros mapeados a must-haves del JD, con métrica.
-3. **Empresa** — por qué *esta* empresa/rol (research real).
-4. **Cierre** — CTA breve + gracias.
+1. **Hook** - gancho específico (logro o conexión real).
+2. **Fit** - 1-2 logros mapeados a must-haves del JD, con métrica.
+3. **Empresa** - por qué *esta* empresa/rol (research real).
+4. **Cierre** - CTA breve + gracias.
 
 ## Estructura (mensaje corto / LinkedIn / email)
 

package/skills/cv-ats-export/SKILL.md

@@ -1,22 +1,23 @@
 ---
 name: cv-ats-export
-description: "Export a Markdown CV to an ATS-compliant PDF (single column, selectable text, A4) using headless Chrome/Edge — no Obsidian or design tool needed. Triggers: exportar CV a PDF, generar PDF del CV, CV ATS PDF, convertir CV markdown a PDF, imprimir CV."
+description: "Export a Markdown CV to an ATS-compliant PDF (single column, selectable text, A4) using headless Chrome/Edge - no Obsidian or design tool needed. Triggers: exportar CV a PDF, generar PDF del CV, CV ATS PDF, convertir CV markdown a PDF, imprimir CV."
 argument-hint: "[cv-name | all] [--workspace <path>]"
 category: "career"
 user-invocable: true
 allowed-tools: Read, Glob, Grep, Bash, Write
 ---
 
-# CV ATS Export — Markdown → PDF ATS-compliant
+# CV ATS Export - Markdown → PDF ATS-compliant
 
-Convertís un CV en Markdown a un **PDF que pasa filtros ATS**: una sola columna, texto seleccionable (no imagen), A4, tipografía limpia. Usa Chrome o Edge en modo headless — no requiere Obsidian, LaTeX ni herramienta de diseño.
+Convertís un CV en Markdown a un **PDF que pasa filtros ATS**: una sola columna, texto seleccionable (no imagen), A4, tipografía limpia. Usa Chrome o Edge en modo headless - no requiere Obsidian, LaTeX ni herramienta de diseño.
 
-El motor es `scripts/cv_export.py`. La primera vez, instalalo en el workspace (`tools/cv_export.py`) para que quede versionado con los CVs; o corrélo directo desde el skill.
+El motor es `scripts/cv_export.py` + `scripts/verify_pdf.py` (verificación ATS post-export). La primera vez, instalá ambos en el workspace (`tools/`) para que queden versionados con los CVs; o corrélos directo desde el skill (van juntos: `cv_export` importa `verify_pdf` desde su mismo directorio).
 
 ## Requisitos
 
 - **Python 3** en el PATH.
 - **Google Chrome o Microsoft Edge** instalado (el script autodetecta rutas estándar en Windows/macOS/Linux).
+- **`pypdf` (opcional)** para la verificación ATS automática: `pip install pypdf`. Sin pypdf el export funciona igual y reporta `[ATS SKIP]`.
 - CVs en Markdown dentro de `02-CVs/` del career workspace, con headers (`# Nombre`, `## Sección`), bullets y `---` como separadores.
 
 ## Uso
@@ -30,6 +31,9 @@ python scripts/cv_export.py "CV - Acme - Backend"
 
 # Workspace explícito
 python scripts/cv_export.py --workspace ./career-workspace "CV - Acme - Backend"
+
+# Saltar la verificación ATS
+python scripts/cv_export.py --no-verify "CV - Acme - Backend"
 ```
 
 El script:
@@ -37,6 +41,15 @@ El script:
 2. Lee el `.md`, separa frontmatter, convierte el cuerpo a HTML con el CSS ATS embebido.
 3. Imprime a PDF con Chrome/Edge headless (`--headless=new --print-to-pdf`).
 4. Guarda en `02-CVs/exports/`. El nombre sale del frontmatter `archivo_pdf` si existe; si no, del nombre del `.md` (sin el prefijo `CV - `).
+5. **Verifica la capa de texto ATS** con `verify_pdf.py` (pypdf): texto extraíble (mínimo 200 chars), que el nombre del H1 y el email sobrevivan al export, y avisa si supera las 2 páginas. Reporta `[ATS OK]` / `[ATS WARN]` / `[ATS SKIP]`; nunca rompe el export. Se salta con `--no-verify`.
+
+Los dos `[ATS SKIP]` son distintos: `verify_pdf.py not found next to cv_export.py` significa que copiaste solo un archivo al workspace (instalar pypdf no cambia nada); `pypdf not installed` es el que se resuelve con `pip install pypdf`.
+
+`verify_pdf.py` también funciona standalone para chequear cualquier PDF:
+
+```bash
+python scripts/verify_pdf.py "02-CVs/exports/Backend.pdf" --min-chars 200 --max-pages 2 --contains "Jane Doe"
+```
 
 ## Markdown soportado
 
@@ -51,12 +64,12 @@ El script:
 
 ## Personalizar el estilo
 
-El CSS ATS está en la constante `CSS` dentro de `scripts/cv_export.py` (tamaño de fuente, márgenes A4, estilos de h1/h2). Editá ahí si querés ajustar tipografía o espaciado — mantené una sola columna y texto real.
+El CSS ATS está en la constante `CSS` dentro de `scripts/cv_export.py` (tamaño de fuente, márgenes A4, estilos de h1/h2). Editá ahí si querés ajustar tipografía o espaciado - mantené una sola columna y texto real.
 
 ## Reglas
 
 - No metas tablas, columnas ni imágenes en el CV: rompen el parseo ATS.
-- Verificá el PDF generado: que el texto sea seleccionable y entre en 1–2 páginas.
+- Atendé los `[ATS WARN]`: un PDF-imagen o sin el nombre/email extraíbles no pasa filtros ATS. Usá `--no-verify` solo para iterar rápido sobre el layout.
 - No commitees PDFs con datos sensibles a un remoto sin confirmación.
 
 ## Handoff

package/skills/cv-ats-export/scripts/cv_export.py

@@ -8,6 +8,12 @@ Usage (from anywhere; the script locates the career workspace):
     python cv_export.py                          # all base CVs (CV - *.md)
     python cv_export.py "CV - Acme - Backend"    # a single CV (with/without .md)
     python cv_export.py --workspace ./career-workspace "CV - Acme"
+    python cv_export.py --no-verify ...          # skip the ATS verification
+
+After exporting, the PDF text layer is verified with verify_pdf.py (pypdf):
+extractable text, the H1 name and the email must survive the export, and a
+warning is printed above 2 pages. Reports [ATS OK] / [ATS WARN]; if pypdf is
+not installed, [ATS SKIP] without breaking the export.
 
 Workspace resolution order:
     1. --workspace <path> flag
@@ -204,7 +210,41 @@ def find_browser():
     return None
 
 
-def export(md_path, browser, cv_dir, export_dir):
+def verify_ats(pdf_path, body):
+    """Verify the exported PDF's text layer. Never breaks the export: reports."""
+    # Two distinct failures, two distinct messages: verify_pdf.py missing next to
+    # this script (installing pypdf would not help) vs pypdf missing, which
+    # verify_pdf only imports when it actually reads the PDF.
+    try:
+        from verify_pdf import verify
+    except ImportError:
+        print(f"        [ATS SKIP] verify_pdf.py not found next to {Path(__file__).name}")
+        return
+
+    contains = []
+    m = re.search(r"^#\s+(.+)$", body, re.MULTILINE)
+    if m:
+        contains.append(re.sub(r"\*|\[|\]\([^)]*\)", "", m.group(1)).strip())
+    m = re.search(r"[\w.+-]+@[\w-]+\.[\w.]+", body)
+    if m:
+        contains.append(m.group(0))
+
+    try:
+        problems = verify(pdf_path, min_chars=200, contains=contains, max_pages=2)
+    except ImportError:
+        print("        [ATS SKIP] pypdf not installed (pip install pypdf)")
+        return
+    except Exception as e:
+        problems = [f"could not verify: {e}"]
+
+    if problems:
+        for p in problems:
+            print(f"        [ATS WARN] {p}")
+    else:
+        print("        [ATS OK] text layer verified (name + email extractable)")
+
+
+def export(md_path, browser, cv_dir, export_dir, check_ats=True):
     text = md_path.read_text(encoding="utf-8")
     fm, body = strip_frontmatter(text)
 
@@ -252,12 +292,15 @@ def export(md_path, browser, cv_dir, export_dir):
 
     size_kb = pdf_path.stat().st_size // 1024
     print(f"[OK] {md_path.name} -> {pdf_path} ({size_kb} KB)")
+    if check_ats:
+        verify_ats(pdf_path, body)
     return True
 
 
 def parse_args(argv):
     workspace = None
     names = []
+    check_ats = True
     i = 0
     while i < len(argv):
         a = argv[i]
@@ -266,14 +309,16 @@ def parse_args(argv):
             workspace = argv[i] if i < len(argv) else None
         elif a.startswith("--workspace="):
             workspace = a.split("=", 1)[1]
+        elif a == "--no-verify":
+            check_ats = False
         else:
             names.append(a)
         i += 1
-    return workspace, names
+    return workspace, names, check_ats
 
 
 def main():
-    workspace, names = parse_args(sys.argv[1:])
+    workspace, names, check_ats = parse_args(sys.argv[1:])
 
     browser = find_browser()
     if not browser:
@@ -298,7 +343,7 @@ def main():
         print(f"[ERROR] No CVs (CV - *.md) found in {cv_dir}.")
         sys.exit(1)
 
-    ok = all([export(t, browser, cv_dir, export_dir) for t in targets])
+    ok = all([export(t, browser, cv_dir, export_dir, check_ats) for t in targets])
     sys.exit(0 if ok else 1)