init

2026-07-13 10:14:00 +03:00
commit d295c7f139
@@ -0,0 +1,95 @@
+# Batch Processing
+
+Process multiple job offers in parallel via headless workers. Each worker runs the full evaluation pipeline (A-F report + PDF + tracker line) autonomously. See the **Headless / Batch Mode** table in `AGENTS.md` for the correct command per CLI.
+
+## Quick Start
+
+1. **Add offers** to `batch-input.tsv` (tab-separated: `id`, `url`, `source`, `notes`):
+
+   ```tsv
+   id	url	source	notes
+   1	https://jobs.example.com/role-a	LinkedIn	
+   2	https://greenhouse.io/company/role-b	Greenhouse	priority
+   ```
+
+2. **Dry run** to preview what will be processed:
+
+   ```bash
+   ./batch/batch-runner.sh --dry-run
+   ```
+
+3. **Run the batch**:
+
+   ```bash
+   ./batch/batch-runner.sh
+   ```
+
+4. **Results** are automatically merged into `data/applications.md`, processed offers are reconciled out of the `data/pipeline.md` inbox, and integrity is verified with `verify-pipeline.mjs` at the end of the run.
+
+## Options
+
+| Flag | Default | Description |
+|------|---------|-------------|
+| `--parallel N` | `1` | Number of concurrent headless workers |
+| `--dry-run` | off | Preview pending offers without processing |
+| `--retry-failed` | off | Only retry offers marked as `failed` in state |
+| `--resume-paused` | off | Resume offers paused after a Claude session/rate limit |
+| `--start-from N` | `0` | Skip offers with ID below N |
+| `--limit N` | `0` | Max number of offers to process in this run (0 = no limit) |
+| `--max-retries N` | `2` | Max retry attempts per offer before giving up |
+| `--rate-limit-sleep N` | `300` | Seconds to wait before retrying a transient rate-limited worker; use `0` to pause the batch immediately |
+
+## Directory Layout
+
+```
+batch/
+  batch-runner.sh          # Orchestrator script
+  batch-prompt.md          # Prompt template sent to each worker
+  batch-input.tsv          # Input offers (you create this)
+  batch-state.tsv          # Processing state (auto-managed, resumable)
+  logs/                    # Per-offer worker logs ({report_num}-{id}.log)
+  tracker-additions/       # TSV lines produced by workers
+    merged/                # TSVs already merged into applications.md
+```
+
+## How It Works
+
+1. **batch-runner.sh** reads `batch-input.tsv` and `batch-state.tsv` to determine which offers need processing.
+2. For each pending offer, it assigns a report number and launches a headless worker with `batch-prompt.md` as the system prompt (placeholders like `{{URL}}`, `{{REPORT_NUM}}` are resolved).
+3. Each worker evaluates the offer, writes a report to `reports/`, generates a PDF to `output/`, and writes a tracker TSV to `tracker-additions/`.
+4. After all workers finish, batch-runner calls `merge-tracker.mjs` to merge TSVs into `data/applications.md`, `reconcile-pipeline.mjs` to move processed offers out of the `data/pipeline.md` inbox, and `verify-pipeline.mjs` to check integrity.
+
+## Tracker Merge
+
+Workers write one TSV per offer to `batch/tracker-additions/`. The merge script (`npm run merge`) handles:
+
+- Deduplication by company + role fuzzy match and report number
+- Column order conversion (TSV has status before score; applications.md has score before status)
+- In-place updates when a re-evaluation scores higher than the existing entry
+- Moving processed TSVs to `tracker-additions/merged/`
+
+Run `npm run merge` manually if you need to merge outside of a batch run.
+
+## Pipeline Reconcile
+
+Batch mode reads offers from `batch-input.tsv`, but the `data/pipeline.md` inbox is a separate list. Without reconciliation, an offer evaluated by a batch run stays in the pipeline "Pendientes" section and gets surfaced again on the next scan or `/career-ops pipeline` run -- producing duplicate reports.
+
+`reconcile-pipeline.mjs` (run as `npm run reconcile`) closes that gap: after the tracker merge, every `completed` or `skipped` offer in `batch-state.tsv` whose URL is still in pipeline "Pendientes" is moved to "Procesadas" with its report link and score (entries without a report file on disk are left in place). It is idempotent -- safe to run after every batch, or manually.
+
+## Resumability
+
+`batch-state.tsv` tracks the status of every offer (`pending`, `processing`, `completed`, `failed`, `skipped`, `rate_limited`, `paused_rate_limit`). If the batch is interrupted, re-running `batch-runner.sh` picks up where it left off -- completed offers are skipped automatically. `rate_limited` is a non-completed state used while the runner waits before retrying, so interrupted rate-limited jobs are eligible on the next normal run.
+
+`paused_rate_limit` is different: it means a worker hit a Claude session/usage limit, so the runner stopped scheduling new offers and preserved the retry count. Resume those rows explicitly after the limit resets:
+
+```bash
+./batch/batch-runner.sh --resume-paused
+```
+
+A PID-based lock file (`batch-runner.pid`) prevents concurrent batch runs. If a previous run crashed, the stale lock is detected and removed automatically.
+
+## Prerequisites
+
+- Your CLI in PATH (see **Headless / Batch Mode** table in `AGENTS.md`)
+- Node.js >= 18, Playwright chromium installed (`npm run doctor` to verify)
+- `batch-input.tsv` with at least one offer
@@ -0,0 +1,11 @@
+id	url	source	notes
+1	https://getmatch.ru/vacancies/34414-starshii-frontend-razrabotchik?s=offers	Javascript jobs (Vue/React/Angular) — Telegram	Старший frontend-разработчик | note: via Telegram: https://t.me/s/jsdevjob/2410
+2	https://hirehi.ru/development/frontend-razrabotchik-55966	Javascript jobs (Vue/React/Angular) — Telegram	Middle Frontend | note: via Telegram: https://t.me/s/jsdevjob/2413
+3	https://hh.ru/vacancy/133613060?query=frontend&hhtmFrom=vacancy_search_list	Javascript jobs (Vue/React/Angular) — Telegram	Frontend-разработчик | note: via Telegram: https://t.me/s/jsdevjob/2416
+4	https://hh.ru/vacancy/133056126?query=frontend&hhtmFrom=vacancy_search_list	Javascript jobs (Vue/React/Angular) — Telegram	Frontend developer (React) | note: via Telegram: https://t.me/s/jsdevjob/2419
+5	https://getmatch.ru/vacancies/35037-frontend-software-engineer-bootcamp?s=offers	Javascript jobs (Vue/React/Angular) — Telegram	Frontend Software Engineer (Bootcamp) | note: via Telegram: https://t.me/s/jsdevjob/2423
+6	https://career.habr.com/vacancies/1000167021	Javascript jobs (Vue/React/Angular) — Telegram	Ведущий фронтенд-разработчик (React) | note: via Telegram: https://t.me/s/jsdevjob/2425
+7	https://hh.ru/vacancy/119867578?query=frontend&hhtmFrom=vacancy_search_list	Javascript jobs (Vue/React/Angular) — Telegram	Frontend-разработчик React (junior) | note: via Telegram: https://t.me/s/jsdevjob/2429
+8	https://t.me/s/runello_rus_javascript/1443	SiteAnalyzer	Frontend на React
+9	https://t.me/s/runello_rus_javascript/1445	JavaScript Job Offers — Telegram	Fullstack React Developer
+10	https://t.me/s/runello_rus_javascript/1448	ЗАО КонсультантПлюс	Веб разработчик Vue/Js
@@ -0,0 +1,458 @@
+# career-ops Batch Worker — Evaluación Completa + PDF + Tracker Line
+
+Eres un worker de evaluación de ofertas de empleo for the candidate (read name from config/profile.yml). Recibes una oferta (URL + JD text) y produces:
+
+1. Evaluación completa A-G (report .md)
+2. PDF personalizado ATS-optimizado
+3. Línea de tracker para merge posterior
+
+**IMPORTANTE**: Este prompt es self-contained. Tienes TODO lo necesario aquí. No dependes de ningún otro skill ni sistema.
+
+---
+
+## Fuentes de Verdad (LEER antes de evaluar)
+
+| Archivo | Ruta absoluta | Cuándo |
+|---------|---------------|--------|
+| cv.md | `cv.md (project root)` | SIEMPRE |
+| _profile.md | `modes/_profile.md (if exists)` | SIEMPRE (user customizations: archetypes, role_shape, location policy, comp targets) |
+| profile.yml | `config/profile.yml (if exists)` | SIEMPRE (candidate identity, comp range, role_shape rules) |
+| llms.txt | `llms.txt (if exists)` | SIEMPRE |
+| article-digest.md | `article-digest.md (project root)` | SIEMPRE (proof points) |
+| i18n.ts | `i18n.ts (if exists, optional)` | Solo entrevistas/deep |
+| cv-template.html | `templates/cv-template.html` | Para PDF |
+| generate-pdf.mjs | `generate-pdf.mjs` | Para PDF |
+
+**REGLA: NUNCA escribir en cv.md ni i18n.ts.** Son read-only.
+**REGLA: NUNCA hardcodear métricas.** Leerlas de cv.md + article-digest.md en el momento.
+**REGLA: Para métricas de artículos, article-digest.md prevalece sobre cv.md.** cv.md puede tener números más antiguos — es normal.
+**REGLA: Antes de evaluar, cargar `modes/_profile.md` y `config/profile.yml` si existen.** Contienen las preferencias del candidato Y reglas concretas de scoring que **sobrescriben** los defaults del sistema.
+
+Tipos de patrones que estos archivos pueden incluir:
+- **Caps de bloque** — ej: "cap Block A at 3.0/5 if title contains 'Lead'/'Head'/'Principal'"
+- **Overrides de recomendación** — ej: "force SKIP if comp ceiling below $120K" o "force SKIP if role_shape signals broad ownership"
+- **Scoring por dimensión** — ej: "Remote: full credit on remote-first; score 2.0 on full on-site outside [region]"
+- **Framing adaptativo por archetype** — mappings entre arquetipos detectados y proof points a priorizar
+
+Aplicación durante la evaluación A-G:
+- **Bloque A:** aplicar caps de role-shape ANTES de calcular el score del bloque
+- **Bloques B-D:** aplicar adaptive framing por archetype y reglas de dimension scoring (location, comp, etc.)
+- **Bloque F:** aplicar recommendation overrides (SKIP forzado, etc.) — `_profile.md` puede convertir un score técnicamente alto en un SKIP por shape o por comp
+
+**En conflicto, las reglas de `_profile.md` ganan sobre los defaults de `_shared.md`.** Esto es intencional: `_profile.md` es la capa de personalización del usuario.
+
+---
+
+## Placeholders (sustituidos por el orquestador)
+
+| Placeholder | Descripción |
+|-------------|-------------|
+| `{{URL}}` | URL de la oferta |
+| `{{JD_FILE}}` | Ruta al archivo con el texto del JD |
+| `{{REPORT_NUM}}` | Número de report (3 dígitos, zero-padded: 001, 002...) |
+| `{{DATE}}` | Fecha actual YYYY-MM-DD |
+| `{{ID}}` | ID único de la oferta en batch-input.tsv |
+
+---
+
+## Pipeline (ejecutar en orden)
+
+### Paso 1 — Obtener JD
+
+1. Lee el archivo JD en `{{JD_FILE}}`
+2. Si el archivo está vacío o no existe, intenta obtener el JD desde `{{URL}}` con WebFetch
+3. Si ambos fallan, reporta error y termina
+
+### Paso 2 — Evaluación A-G
+
+Read `cv.md`. Ejecuta TODOS los bloques:
+
+#### Paso 0 — Detección de Arquetipo
+
+Clasifica la oferta en uno de los 6 arquetipos. Si es híbrido, indica los 2 más cercanos.
+
+**Los 6 arquetipos (todos igual de válidos):**
+
+| Arquetipo | Ejes temáticos | Qué compran |
+|-----------|----------------|-------------|
+| **AI Platform / LLMOps Engineer** | Evaluation, observability, reliability, pipelines | Alguien que ponga AI en producción con métricas |
+| **Agentic Workflows / Automation** | HITL, tooling, orchestration, multi-agent | Alguien que construya sistemas de agentes fiables |
+| **Technical AI Product Manager** | GenAI/Agents, PRDs, discovery, delivery | Alguien que traduzca negocio → producto AI |
+| **AI Solutions Architect** | Hyperautomation, enterprise, integrations | Alguien que diseñe arquitecturas AI end-to-end |
+| **AI Forward Deployed Engineer** | Client-facing, fast delivery, prototyping | Alguien que entregue soluciones AI a clientes rápido |
+| **AI Transformation Lead** | Change management, adoption, org enablement | Alguien que lidere el cambio AI en una organización |
+
+**Framing adaptativo:**
+
+> **Las métricas concretas se leen de `cv.md` + `article-digest.md` en cada evaluación. NUNCA hardcodear números aquí.**
+
+| Si el rol es... | Emphasize about the candidate... | Fuentes de proof points |
+|-----------------|--------------------------|--------------------------|
+| Platform / LLMOps | Builder de sistemas en producción, observability, evals, closed-loop | article-digest.md + cv.md |
+| Agentic / Automation | Orquestación multi-agente, HITL, reliability, cost | article-digest.md + cv.md |
+| Technical AI PM | Product discovery, PRDs, métricas, stakeholder mgmt | cv.md + article-digest.md |
+| Solutions Architect | Diseño de sistemas, integrations, enterprise-ready | article-digest.md + cv.md |
+| Forward Deployed Engineer | Fast delivery, client-facing, prototype → prod | cv.md + article-digest.md |
+| AI Transformation Lead | Change management, team enablement, adoption | cv.md + article-digest.md |
+
+**Ventaja transversal**: Enmarcar perfil como **"Technical builder"** que adapta su framing al rol:
+- Para PM: "builder que reduce incertidumbre con prototipos y luego productioniza con disciplina"
+- Para FDE: "builder que entrega fast con observability y métricas desde día 1"
+- Para SA: "builder que diseña sistemas end-to-end con experiencia real en integrations"
+- Para LLMOps: "builder que pone AI en producción con closed-loop quality systems — leer métricas de article-digest.md"
+
+Convertir "builder" en señal profesional, no en "hobby maker". El framing cambia, la verdad es la misma.
+
+#### Bloque A — Resumen del Rol
+
+Tabla con: Arquetipo detectado, Domain, Function, Seniority, Remote, Team size, TL;DR.
+
+#### Bloque B — Match con CV
+
+Read `cv.md`. Tabla con cada requisito del JD mapeado a líneas exactas del CV o keys de i18n.ts.
+
+**Adaptado al arquetipo:**
+- FDE → priorizar delivery rápida y client-facing
+- SA → priorizar diseño de sistemas e integrations
+- PM → priorizar product discovery y métricas
+- LLMOps → priorizar evals, observability, pipelines
+- Agentic → priorizar multi-agent, HITL, orchestration
+- Transformation → priorizar change management, adoption, scaling
+
+Sección de **gaps** con estrategia de mitigación para cada uno:
+1. ¿Es hard blocker o nice-to-have?
+2. Can the candidate demonstrate experiencia adyacente?
+3. ¿Hay un proyecto portfolio que cubra este gap?
+4. Plan de mitigación concreto
+
+#### Bloque C — Nivel y Estrategia
+
+1. **Nivel detectado** en el JD vs **candidate's natural level**
+2. **Plan "vender senior sin mentir"**: frases específicas, logros concretos, founder como ventaja
+3. **Plan "si me downlevelan"**: aceptar si comp justa, review a 6 meses, criterios claros
+
+#### Bloque D — Comp y Demanda
+
+Usar WebSearch para salarios actuales (Glassdoor, Levels.fyi, Blind), reputación comp de la empresa, tendencia demanda. Tabla con datos y fuentes citadas. Si no hay datos, decirlo.
+
+Score de comp (1-5): 5=top quartile, 4=above market, 3=median, 2=slightly below, 1=well below.
+
+#### Bloque E — Plan de Personalización
+
+| # | Sección | Estado actual | Cambio propuesto | Por qué |
+|---|---------|---------------|------------------|---------|
+
+Top 5 cambios al CV + Top 5 cambios a LinkedIn.
+
+#### Bloque F — Plan de Entrevistas
+
+6-10 historias STAR mapeadas a requisitos del JD:
+
+| # | Requisito del JD | Historia STAR | S | T | A | R |
+
+**Selección adaptada al arquetipo.** Incluir también:
+- 1 case study recomendado (cuál proyecto presentar y cómo)
+- Preguntas red-flag y cómo responderlas
+
+#### Bloque G — Posting Legitimacy
+
+Analyze posting signals to assess whether this is a real, active opening.
+
+**Batch mode limitations:** Playwright is not available, so posting freshness signals (exact days posted, apply button state) cannot be directly verified. Mark these as "unverified (batch mode)."
+
+**What IS available in batch mode:**
+1. **Description quality analysis** -- Full JD text is available. Analyze specificity, requirements realism, salary transparency, boilerplate ratio.
+2. **Company hiring signals** -- WebSearch queries for layoff/freeze news (combine with Block D comp research).
+3. **Reposting detection** -- Read `data/scan-history.tsv` to check for prior appearances.
+4. **Role market context** -- Qualitative assessment from JD content.
+
+**Output format:** Same as interactive mode (Assessment tier + Signals table + Context Notes), but with a note that posting freshness is unverified.
+
+**Assessment:** Apply the same three tiers (High Confidence / Proceed with Caution / Suspicious), weighting available signals more heavily. If insufficient signals are available to make a determination, default to "Proceed with Caution" with a note about limited data.
+
+#### Score Global
+
+| Dimensión | Score |
+|-----------|-------|
+| Match con CV | X/5 |
+| Alineación North Star | X/5 |
+| Comp | X/5 |
+| Señales culturales | X/5 |
+| Red flags | -X (si hay) |
+| **Global** | **X/5** |
+
+#### Machine Summary
+
+Create a machine-readable summary from the completed A-G evaluation and global score. This block is for downstream scripts; keep field names exact, use YAML, and do not add prose inside the fence.
+
+```yaml
+company: "{empresa}"
+role: "{rol}"
+score: {X.X}
+legitimacy_tier: "{High Confidence | Proceed with Caution | Suspicious}"
+archetype: "{detectado}"
+final_decision: "{Apply | Consider | Research first | Skip}"
+hard_stops:
+  - "{blocking gap or risk}"
+soft_gaps:
+  - "{non-blocking gap}"
+top_strengths:
+  - "{strength most relevant to this role}"
+risk_level: "{Low | Medium | High}"
+confidence: "{Low | Medium | High}"
+next_action: "{one concrete next step}"
+```
+
+Rules:
+- Use `[]` for `hard_stops`, `soft_gaps`, or `top_strengths` when empty.
+- `score` is numeric only, without `/5`.
+- `final_decision` must reflect the full evaluation, not only the CV match.
+- Do not invent missing data. If confidence is limited, set `confidence: "Low"` and explain the limitation in the human-readable sections.
+
+### Paso 3 — Guardar Report .md
+
+Guardar evaluación completa en:
+```
+reports/{{REPORT_NUM}}-{company-slug}-{{DATE}}.md
+```
+
+Donde `{company-slug}` es el nombre de empresa en lowercase, sin espacios, con guiones.
+
+**Formato del report:**
+
+```markdown
+# Evaluación: {Empresa} — {Rol}
+
+**Fecha:** {{DATE}}
+**Arquetipo:** {detectado}
+**Score:** {X/5}
+**Legitimacy:** {High Confidence | Proceed with Caution | Suspicious}
+**URL:** {URL de la oferta original}
+**PDF:** {output/cv-candidate-{company-slug}-{{DATE}}.pdf if score ≥ the resolved `auto_pdf_score_threshold` from Paso 4, else `not generated — run /career-ops pdf {company-slug} to create on demand`}
+**Batch ID:** {{ID}}
+
+---
+
+## Machine Summary
+
+```yaml
+company: "{empresa}"
+role: "{rol}"
+score: {X.X}
+legitimacy_tier: "{High Confidence | Proceed with Caution | Suspicious}"
+archetype: "{detectado}"
+final_decision: "{Apply | Consider | Research first | Skip}"
+hard_stops:
+  - "{blocking gap or risk}"
+soft_gaps:
+  - "{non-blocking gap}"
+top_strengths:
+  - "{strength most relevant to this role}"
+risk_level: "{Low | Medium | High}"
+confidence: "{Low | Medium | High}"
+next_action: "{one concrete next step}"
+```
+
+## A) Resumen del Rol
+(contenido completo)
+
+## B) Match con CV
+(contenido completo)
+
+## C) Nivel y Estrategia
+(contenido completo)
+
+## D) Comp y Demanda
+(contenido completo)
+
+## E) Plan de Personalización
+(contenido completo)
+
+## F) Plan de Entrevistas
+(contenido completo)
+
+## G) Posting Legitimacy
+(contenido completo)
+
+---
+
+## Keywords extraídas
+(15-20 keywords del JD para ATS)
+```
+
+### Paso 4 — Generar PDF (configurable)
+
+**Gate:** Read `config/profile.yml` → `auto_pdf_score_threshold`. If the key is absent, default to **`3.0`** (the original gate of Path A). This step ONLY runs when the score from Paso 2 is **≥ the resolved threshold**. For everything below it, skip this entire step — the user can generate a tailored PDF on demand later via `/career-ops pdf {company-slug}` using the report from Paso 3 as input.
+
+**Rationale:** Generating a tailored PDF costs ~30–60s per offer (Playwright launch + HTML render) and produces files that often go unused — most roles score 2.x/3.x and never reach application. The `3.0` default matches Path A's original behavior; raise `auto_pdf_score_threshold` (e.g. `4.0`) to pre-generate fewer PDFs, or set `0` to generate one for every offer. Both Path A (`/career-ops pipeline`) and Path B (this batch worker) read the same config key for consistency.
+
+**If score < threshold:**
+- Skip steps 1–14 below.
+- In the report header use: `**PDF:** not generated — run /career-ops pdf {company-slug} to create on demand`.
+- In Paso 5 (tracker line) use `pdf_emoji` = `❌`.
+- In Paso 6 (output JSON) set `"pdf": null`.
+- Done — move to Paso 5.
+
+**If score ≥ threshold**, generate the tailored PDF:
+
+1. Lee `cv.md` + `i18n.ts`
+2. Extrae 15-20 keywords del JD
+3. Detecta idioma del JD → idioma del CV (EN default)
+4. Detecta ubicación empresa → formato papel: US/Canada → `letter`, resto → `a4`
+5. Detecta arquetipo → adapta framing
+6. Reescribe Professional Summary inyectando keywords
+7. Selecciona top 3-4 proyectos más relevantes
+8. Reordena bullets de experiencia por relevancia al JD
+9. Construye competency grid (6-8 keyword phrases)
+10. Inyecta keywords en logros existentes (**NUNCA inventa**)
+11. Genera HTML completo desde template (lee `templates/cv-template.html`)
+12. Escribe HTML a `output/cv-candidate-{company-slug}.html` (NO en /tmp — el HTML registrado es la fuente de regeneración del dashboard)
+13. Ejecuta:
+```bash
+node generate-pdf.mjs \
+  output/cv-candidate-{company-slug}.html \
+  output/cv-candidate-{company-slug}-{{DATE}}.pdf \
+  --format={letter|a4} \
+  --report={{REPORT_NUM}}
+```
+14. Reporta: ruta PDF, nº páginas, % cobertura keywords
+
+On success, in Paso 5 use `pdf_emoji` = `✅` and in Paso 6 set `"pdf"` to the output path.
+
+**Reglas ATS:**
+- Single-column (sin sidebars)
+- Headers estándar: "Professional Summary", "Work Experience", "Education", "Skills", "Certifications", "Projects"
+- Sin texto en imágenes/SVGs
+- Sin info crítica en headers/footers
+- UTF-8, texto seleccionable
+- Keywords distribuidas: Summary (top 5), primer bullet de cada rol, Skills section
+
+**Diseño:**
+- Fonts: Space Grotesk (headings, 600-700) + DM Sans (body, 400-500)
+- Fonts self-hosted: `fonts/`
+- Header: Space Grotesk 24px bold + gradiente cyan→purple 2px + contacto
+- Section headers: Space Grotesk 13px uppercase, color cyan `hsl(187,74%,32%)`
+- Body: DM Sans 11px, line-height 1.5
+- Company names: purple `hsl(270,70%,45%)`
+- Márgenes: 0.6in
+- Background: blanco
+
+**Estrategia keyword injection (ético):**
+- Reformular experiencia real con vocabulario exacto del JD
+- NUNCA añadir skills the candidate doesn't have
+- Ejemplo: JD dice "RAG pipelines" y CV dice "LLM workflows with retrieval" → "RAG pipeline design and LLM orchestration workflows"
+
+**Template placeholders (en cv-template.html):**
+
+| Placeholder | Contenido |
+|-------------|-----------|
+| `{{LANG}}` | `en` o `es` |
+| `{{PAGE_WIDTH}}` | `8.5in` (letter) o `210mm` (A4) |
+| `{{NAME}}` | (from profile.yml) |
+| `{{EMAIL}}` | (from profile.yml) |
+| `{{LINKEDIN_URL}}` | (from profile.yml) |
+| `{{LINKEDIN_DISPLAY}}` | (from profile.yml) |
+| `{{PORTFOLIO_URL}}` | (from profile.yml) |
+| `{{PORTFOLIO_DISPLAY}}` | (from profile.yml) |
+| `{{LOCATION}}` | (from profile.yml) |
+| `{{SECTION_SUMMARY}}` | Professional Summary / Resumen Profesional |
+| `{{SUMMARY_TEXT}}` | Summary personalizado con keywords |
+| `{{SECTION_COMPETENCIES}}` | Core Competencies / Competencias Core |
+| `{{COMPETENCIES}}` | `<span class="competency-tag">keyword</span>` × 6-8 |
+| `{{SECTION_EXPERIENCE}}` | Work Experience / Experiencia Laboral |
+| `{{EXPERIENCE}}` | HTML de cada trabajo con bullets reordenados |
+| `{{SECTION_PROJECTS}}` | Projects / Proyectos |
+| `{{PROJECTS}}` | HTML de top 3-4 proyectos |
+| `{{SECTION_EDUCATION}}` | Education / Formación |
+| `{{EDUCATION}}` | HTML de educación |
+| `{{SECTION_CERTIFICATIONS}}` | Certifications / Certificaciones |
+| `{{CERTIFICATIONS}}` | HTML de certificaciones |
+| `{{SECTION_SKILLS}}` | Skills / Competencias |
+| `{{SKILLS}}` | HTML de skills |
+
+### Paso 5 — Tracker Line
+
+Escribir una línea TSV a:
+```
+batch/tracker-additions/{{ID}}.tsv
+```
+
+Formato TSV (una sola línea, sin header, 9 columnas tab-separated):
+```
+{next_num}\t{{DATE}}\t{empresa}\t{rol}\t{status}\t{score}/5\t{pdf_emoji}\t[{{REPORT_NUM}}](reports/{{REPORT_NUM}}-{company-slug}-{{DATE}}.md)\t{nota_1_frase}
+```
+
+**Columnas TSV (orden exacto):**
+
+| # | Campo | Tipo | Ejemplo | Validación |
+|---|-------|------|---------|------------|
+| 1 | num | int | `647` | Secuencial, max existente + 1 |
+| 2 | date | YYYY-MM-DD | `2026-03-14` | Fecha de evaluación |
+| 3 | company | string | `Datadog` | Nombre corto de empresa |
+| 4 | role | string | `Staff AI Engineer` | Título del rol |
+| 5 | status | canonical | `Evaluada` | DEBE ser canónico (ver states.yml) |
+| 6 | score | X.XX/5 | `4.55/5` | O `N/A` si no evaluable |
+| 7 | pdf | emoji | `✅` o `❌` | Si se generó PDF |
+| 8 | report | md link | `[647](reports/647-...)` | Link root-relative; merge-tracker.mjs lo normaliza relativo al tracker (ej. `../reports/...`, #760) |
+| 9 | notes | string | `APPLY HIGH...` | Resumen 1 frase |
+
+**IMPORTANTE:** El orden TSV tiene status ANTES de score (col 5→status, col 6→score). En applications.md el orden es inverso (col 5→score, col 6→status). merge-tracker.mjs maneja la conversión.
+
+**Estados canónicos válidos:** `Evaluada`, `Aplicado`, `Respondido`, `Entrevista`, `Oferta`, `Rechazado`, `Descartado`, `NO APLICAR`
+
+Donde `{next_num}` se calcula leyendo la última línea de `data/applications.md`.
+
+### Paso 6 — Output final
+
+Al terminar, imprime por stdout un resumen JSON para que el orquestador lo parsee:
+
+```json
+{
+  "status": "completed",
+  "id": "{{ID}}",
+  "report_num": "{{REPORT_NUM}}",
+  "company": "{empresa}",
+  "role": "{rol}",
+  "score": {score_num},
+  "legitimacy": "{High Confidence|Proceed with Caution|Suspicious}",
+  "pdf": "{ruta_pdf}",
+  "report": "{ruta_report}",
+  "error": null
+}
+```
+
+Si algo falla:
+```json
+{
+  "status": "failed",
+  "id": "{{ID}}",
+  "report_num": "{{REPORT_NUM}}",
+  "company": "{empresa_o_unknown}",
+  "role": "{rol_o_unknown}",
+  "score": null,
+  "pdf": null,
+  "report": "{ruta_report_si_existe}",
+  "error": "{descripción_del_error}"
+}
+```
+
+---
+
+## Reglas Globales
+
+### NUNCA
+1. Inventar experiencia o métricas
+2. Modificar cv.md, i18n.ts ni archivos del portfolio
+3. Compartir el teléfono en mensajes generados
+4. Recomendar comp por debajo de mercado
+5. Generar PDF sin leer primero el JD
+6. Usar corporate-speak
+
+### SIEMPRE
+1. Leer cv.md, llms.txt y article-digest.md antes de evaluar
+2. Detectar el arquetipo del rol y adaptar el framing
+3. Citar líneas exactas del CV cuando haga match
+4. Usar WebSearch para datos de comp y empresa
+5. Generar contenido en el idioma del JD (EN default)
+6. Ser directo y accionable — sin fluff
+7. Cuando generes texto en inglés (PDF summaries, bullets, STAR stories), usa inglés nativo de tech: frases cortas, verbos de acción, sin passive voice innecesaria, sin "in order to" ni "utilized"
@@ -0,0 +1,932 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# career-ops batch runner — standalone orchestrator for claude -p workers
+# Reads batch-input.tsv, delegates each offer to a claude -p worker,
+# tracks state in batch-state.tsv for resumability.
+#
+# NOTE: This script is Claude Code-specific. It uses claude -p with
+# --dangerously-skip-permissions and --append-system-prompt-file flags
+# that are not available in other CLIs. Multi-CLI support is out of scope
+# for now — contributions welcome.
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+PROJECT_DIR="$(cd "$SCRIPT_DIR/.." && pwd)"
+BATCH_DIR="$SCRIPT_DIR"
+INPUT_FILE="$BATCH_DIR/batch-input.tsv"
+STATE_FILE="$BATCH_DIR/batch-state.tsv"
+PROMPT_FILE="$BATCH_DIR/batch-prompt.md"
+LOGS_DIR="$BATCH_DIR/logs"
+TRACKER_DIR="$BATCH_DIR/tracker-additions"
+REPORTS_DIR="$PROJECT_DIR/reports"
+APPLICATIONS_FILE="$PROJECT_DIR/data/applications.md"
+LOCK_FILE="$BATCH_DIR/batch-runner.pid"
+PAUSE_FILE="$BATCH_DIR/batch-runner.paused"
+STATE_LOCK_DIR="$BATCH_DIR/.batch-state.lock"
+STATE_LOCK_PID_FILE="$STATE_LOCK_DIR/pid"
+STATE_LOCK_TIMEOUT_SECONDS=30
+MAIN_PID="${BASHPID:-$$}"
+
+# Defaults
+PARALLEL=1
+DRY_RUN=false
+RETRY_FAILED=false
+RESUME_PAUSED=false
+START_FROM=0
+MAX_RETRIES=2
+MIN_SCORE=0
+SKIP_PDF=false
+MODEL=""  # empty = let claude -p use the Claude Max default
+RATE_LIMIT_SLEEP=300
+BATCH_PAUSED=false
+STATUS_ONLY=false
+WATCH_MODE=false
+LIMIT=0
+
+# Return success for non-negative integer or decimal strings.
+is_decimal_number() {
+  [[ "$1" =~ ^[0-9]+([.][0-9]+)?$ ]]
+}
+
+usage() {
+  cat <<'USAGE'
+career-ops batch runner — process job offers in batch via claude -p workers
+Uses your default Claude model (Claude Max subscription).
+
+Usage: batch-runner.sh [OPTIONS]
+
+Options:
+  --parallel N         Number of parallel workers (default: 1)
+  --dry-run            Show what would be processed, don't execute
+  --retry-failed       Only retry offers marked as "failed" in state
+  --resume-paused      Resume offers paused by a Claude session/rate limit
+  --start-from N       Start from offer ID N (skip earlier IDs)
+  --limit N            Max number of offers to process in this run
+  --max-retries N      Max retry attempts per offer (default: 2)
+  --min-score N        Skip PDF/tracker for offers scoring below N (default: 0 = off)
+  --skip-pdf           Skip PDF generation entirely (write ❌ in tracker PDF column)
+  --rate-limit-sleep N Seconds to wait before retrying a rate-limited worker
+                       (default: 300)
+  --model NAME         Claude model passed to `claude -p --model` (default:
+                       unset = Claude Max default). Use a cheaper model for
+                       large batches, e.g. `--model claude-sonnet-4-6`.
+  --status             Show batch progress and a per-job table, then exit
+  --watch              Live-refresh progress until the run completes
+  -h, --help           Show this help
+
+Files:
+  batch-input.tsv      Input offers (id, url, source, notes)
+  batch-state.tsv      Processing state (auto-managed)
+  batch-prompt.md      Prompt template for workers
+  logs/                Per-offer logs
+  tracker-additions/   Tracker lines for post-batch merge
+
+Examples:
+  # Dry run to see pending offers
+  ./batch-runner.sh --dry-run
+
+  # Process all pending
+  ./batch-runner.sh
+
+  # Retry only failed offers
+  ./batch-runner.sh --retry-failed
+
+  # Process 2 at a time starting from ID 10
+  ./batch-runner.sh --parallel 2 --start-from 10
+USAGE
+}
+
+# Parse arguments
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+    --parallel) PARALLEL="$2"; shift 2 ;;
+    --dry-run) DRY_RUN=true; shift ;;
+    --retry-failed) RETRY_FAILED=true; shift ;;
+    --resume-paused) RESUME_PAUSED=true; shift ;;
+    --start-from) START_FROM="$2"; shift 2 ;;
+    --limit) LIMIT="$2"; shift 2 ;;
+    --max-retries) MAX_RETRIES="$2"; shift 2 ;;
+    --min-score) MIN_SCORE="$2"; shift 2 ;;
+    --skip-pdf) SKIP_PDF=true; shift ;;
+    --rate-limit-sleep)
+      [[ $# -ge 2 ]] || { echo "ERROR: --rate-limit-sleep requires an argument"; exit 1; }
+      RATE_LIMIT_SLEEP="$2"
+      shift 2
+      ;;
+    --model) MODEL="$2"; shift 2 ;;
+    --status) STATUS_ONLY=true; shift ;;
+    --watch) WATCH_MODE=true; shift ;;
+    -h|--help) usage; exit 0 ;;
+    *) echo "Unknown option: $1"; usage; exit 1 ;;
+  esac
+done
+
+if ! [[ "$RATE_LIMIT_SLEEP" =~ ^[0-9]+$ ]]; then
+  echo "ERROR: --rate-limit-sleep must be a non-negative integer (seconds)."
+  exit 1
+fi
+
+if ! is_decimal_number "$MIN_SCORE"; then
+  echo "ERROR: --min-score must be a non-negative number."
+  exit 1
+fi
+
+if ! [[ "$LIMIT" =~ ^[0-9]+$ ]]; then
+  echo "ERROR: --limit must be a non-negative integer."
+  exit 1
+fi
+
+# Lock file to prevent double execution
+acquire_lock() {
+  if [[ -f "$LOCK_FILE" ]]; then
+    local old_pid
+    old_pid=$(cat "$LOCK_FILE")
+    if kill -0 "$old_pid" 2>/dev/null; then
+      echo "ERROR: Another batch-runner is already running (PID $old_pid)"
+      echo "If this is stale, remove $LOCK_FILE"
+      exit 1
+    else
+      echo "WARN: Stale lock file found (PID $old_pid not running). Removing."
+      rm -f "$LOCK_FILE"
+    fi
+  fi
+  echo "$MAIN_PID" > "$LOCK_FILE"
+}
+
+release_lock() {
+  if [[ "${BASHPID:-$$}" != "$MAIN_PID" ]]; then
+    return
+  fi
+  rm -f "$LOCK_FILE"
+}
+
+trap release_lock EXIT
+
+# Validate prerequisites
+check_prerequisites() {
+  if [[ ! -f "$INPUT_FILE" ]]; then
+    echo "ERROR: $INPUT_FILE not found. Add offers first."
+    exit 1
+  fi
+
+  if [[ ! -f "$PROMPT_FILE" ]]; then
+    echo "ERROR: $PROMPT_FILE not found."
+    exit 1
+  fi
+
+  if ! command -v claude &>/dev/null; then
+    echo "ERROR: 'claude' CLI not found in PATH."
+    exit 1
+  fi
+
+  mkdir -p "$LOGS_DIR" "$TRACKER_DIR" "$REPORTS_DIR"
+}
+
+# Status/watch mode only needs prior batch state, not worker prerequisites.
+check_status_prerequisites() {
+  if [[ ! -f "$STATE_FILE" ]]; then
+    echo "No state file found at $STATE_FILE"
+    exit 0
+  fi
+}
+
+# Initialize state file if it doesn't exist
+init_state() {
+  if [[ ! -f "$STATE_FILE" ]]; then
+    printf 'id\turl\tstatus\tstarted_at\tcompleted_at\treport_num\tscore\terror\tretries\n' > "$STATE_FILE"
+  fi
+}
+
+acquire_state_lock() {
+  if [[ "${STATE_LOCK_DISABLED:-0}" -eq 1 ]]; then
+    return 0
+  fi
+
+  local waited=0
+  local max_waits=$((STATE_LOCK_TIMEOUT_SECONDS * 10))
+
+  while true; do
+    if mkdir "$STATE_LOCK_DIR" 2>/dev/null; then
+      if printf '%s\n' "${BASHPID:-$$}" > "$STATE_LOCK_PID_FILE"; then
+        STATE_LOCK_OWNED=1
+        return 0
+      fi
+      rm -f "$STATE_LOCK_PID_FILE" 2>/dev/null || true
+      rmdir "$STATE_LOCK_DIR" 2>/dev/null || true
+      echo "ERROR: Failed to initialize state lock metadata at $STATE_LOCK_DIR"
+      return 1
+    fi
+
+    if [[ ! -d "$STATE_LOCK_DIR" ]]; then
+      if (( PARALLEL <= 1 )); then
+        echo "WARN: State lock creation failed. Falling back to lock-free operation (single-worker mode)." >&2
+        STATE_LOCK_DISABLED=1
+        STATE_LOCK_OWNED=0
+        return 0
+      fi
+      echo "ERROR: Failed to create state lock directory $STATE_LOCK_DIR"
+      return 1
+    fi
+
+    if [[ -f "$STATE_LOCK_PID_FILE" ]]; then
+      local lock_pid
+      lock_pid=$(cat "$STATE_LOCK_PID_FILE" 2>/dev/null || true)
+      if [[ -n "$lock_pid" ]] && ! kill -0 "$lock_pid" 2>/dev/null; then
+        rm -f "$STATE_LOCK_PID_FILE"
+        if rmdir "$STATE_LOCK_DIR" 2>/dev/null; then
+          echo "WARN: Recovered stale state lock (PID $lock_pid not running)."
+          continue
+        fi
+      fi
+    fi
+
+    if (( waited >= max_waits )); then
+      echo "ERROR: Timed out waiting for state lock at $STATE_LOCK_DIR"
+      echo "If no batch-runner worker is active, remove the stale lock directory."
+      return 1
+    fi
+
+    sleep 0.1
+    ((waited += 1))
+  done
+}
+
+release_state_lock() {
+  if [[ "${STATE_LOCK_OWNED:-0}" -ne 1 ]]; then
+    return
+  fi
+  rm -f "$STATE_LOCK_PID_FILE" 2>/dev/null || true
+  rmdir "$STATE_LOCK_DIR" 2>/dev/null || true
+  STATE_LOCK_OWNED=0
+}
+
+run_with_state_lock() {
+  acquire_state_lock || return $?
+
+  local status=0
+  if "$@"; then
+    status=0
+  else
+    status=$?
+  fi
+
+  release_state_lock
+  return "$status"
+}
+
+# Get status of an offer from state file
+get_status() {
+  local id="$1"
+  if [[ ! -f "$STATE_FILE" ]]; then
+    echo "none"
+    return
+  fi
+  local status
+  status=$(awk -F'\t' -v id="$id" '$1 == id { print $3 }' "$STATE_FILE")
+  echo "${status:-none}"
+}
+
+# Get retry count for an offer
+get_retries() {
+  local id="$1"
+  if [[ ! -f "$STATE_FILE" ]]; then
+    echo "0"
+    return
+  fi
+  local retries
+  retries=$(awk -F'\t' -v id="$id" '$1 == id { print $9 }' "$STATE_FILE")
+  echo "${retries:-0}"
+}
+
+# Calculate next report number.
+# Caller must hold STATE_LOCK_DIR while this runs.
+next_report_num_unlocked() {
+  local max_num=0
+  if [[ -d "$REPORTS_DIR" ]]; then
+    for f in "$REPORTS_DIR"/*.md; do
+      [[ -f "$f" ]] || continue
+      local basename
+      basename=$(basename "$f")
+      local num="${basename%%-*}"
+      num=$((10#$num)) # Remove leading zeros for arithmetic
+      if (( num > max_num )); then
+        max_num=$num
+      fi
+    done
+  fi
+  # Also check state file for assigned report numbers
+  if [[ -f "$STATE_FILE" ]]; then
+    while IFS=$'\t' read -r _ _ _ _ _ rnum _ _ _; do
+      [[ "$rnum" == "report_num" || "$rnum" == "-" || -z "$rnum" ]] && continue
+      local n=$((10#$rnum))
+      if (( n > max_num )); then
+        max_num=$n
+      fi
+    done < "$STATE_FILE"
+  fi
+  printf '%03d' $((max_num + 1))
+}
+
+# Update or insert state for an offer.
+# Caller must hold STATE_LOCK_DIR while this runs.
+update_state_unlocked() {
+  local id="$1" url="$2" status="$3" started="$4" completed="$5" report_num="$6" score="$7" error="$8" retries="$9"
+
+  if [[ ! -f "$STATE_FILE" ]]; then
+    init_state
+  fi
+
+  local tmp="$STATE_FILE.tmp"
+  local found=false
+
+  # Write header
+  head -1 "$STATE_FILE" > "$tmp"
+
+  # Process existing lines
+  while IFS=$'\t' read -r sid surl sstatus sstarted scompleted sreport sscore serror sretries; do
+    [[ "$sid" == "id" ]] && continue  # skip header
+    if [[ "$sid" == "$id" ]]; then
+      printf '%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\n' \
+        "$id" "$url" "$status" "$started" "$completed" "$report_num" "$score" "$error" "$retries" >> "$tmp"
+      found=true
+    else
+      printf '%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\n' \
+        "$sid" "$surl" "$sstatus" "$sstarted" "$scompleted" "$sreport" "$sscore" "$serror" "$sretries" >> "$tmp"
+    fi
+  done < "$STATE_FILE"
+
+  if [[ "$found" == "false" ]]; then
+    printf '%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\n' \
+      "$id" "$url" "$status" "$started" "$completed" "$report_num" "$score" "$error" "$retries" >> "$tmp"
+  fi
+
+  mv "$tmp" "$STATE_FILE"
+}
+
+update_state() {
+  run_with_state_lock update_state_unlocked "$@"
+}
+
+is_rate_limit_log() {
+  local log_file="$1"
+  grep -Eiq '(rate limit|rate_limit|too many requests|429|quota exceeded|try again later|temporarily unavailable)' "$log_file"
+}
+
+is_session_limit_log() {
+  local log_file="$1"
+  grep -Eiq '(session limit|resets [0-9:]+[ap]m|usage limit|limit[[:space:]]+reached)' "$log_file"
+}
+
+mark_paused_rate_limit() {
+  local id="$1" url="$2" started_at="$3" report_num="$4" retries="$5" log_file="$6"
+  local completed_at
+  completed_at=$(date -u +%Y-%m-%dT%H:%M:%SZ)
+  local error_msg
+  error_msg=$(tail -5 "$log_file" 2>/dev/null | tr '\n' ' ' | cut -c1-200 || echo "session/rate limit reached")
+  update_state "$id" "$url" "paused_rate_limit" "$started_at" "$completed_at" "$report_num" "-" "$error_msg" "$retries"
+  printf '%s\t%s\t%s\n' "$id" "$report_num" "$error_msg" > "$PAUSE_FILE"
+  BATCH_PAUSED=true
+}
+
+reserve_report_num_unlocked() {
+  local id="$1" url="$2" started="$3" retries="$4"
+
+  local report_num=""
+  if report_num=$(next_report_num_unlocked); then
+    update_state_unlocked "$id" "$url" "processing" "$started" "-" "$report_num" "-" "-" "$retries"
+  fi
+
+  printf '%s\n' "$report_num"
+}
+
+reserve_report_num() {
+  run_with_state_lock reserve_report_num_unlocked "$@"
+}
+
+# Process a single offer
+process_offer() {
+  local id="$1" url="$2" source="$3" notes="$4"
+
+  local started_at
+  started_at=$(date -u +%Y-%m-%dT%H:%M:%SZ)
+  local retries
+  retries=$(get_retries "$id")
+  local report_num
+  report_num=$(reserve_report_num "$id" "$url" "$started_at" "$retries")
+  local date
+  date=$(date +%Y-%m-%d)
+  # Use mktemp instead of a predictable /tmp path: a fixed name like
+  # /tmp/batch-jd-${id}.txt is guessable, so an attacker on a shared machine
+  # could pre-create it as a symlink and redirect or clobber the write.
+  local jd_file
+  jd_file="$(mktemp "${TMPDIR:-/tmp}/batch-jd-${id}.XXXXXX")"
+
+  echo "--- Processing offer #$id: $url (report $report_num, attempt $((retries + 1)))"
+
+  # Build the prompt with placeholders replaced
+  local prompt
+  if [[ "$SKIP_PDF" == "true" ]]; then
+    prompt="Procesa esta oferta de empleo. Ejecuta el pipeline: evaluación A-F + report .md + tracker line. NO generes PDF; en el tracker escribe ❌ en la columna PDF y en el JSON final establece \"pdf\": null."
+    echo "    ⏭️  --skip-pdf set — skipping PDF generation for #$id ($url)"
+  else
+    prompt="Procesa esta oferta de empleo. Ejecuta el pipeline completo: evaluación A-F + report .md + PDF + tracker line."
+  fi
+  prompt="$prompt URL: $url"
+  prompt="$prompt JD file: $jd_file"
+  prompt="$prompt Report number: $report_num"
+  prompt="$prompt Date: $date"
+  prompt="$prompt Batch ID: $id"
+
+  local log_file="$LOGS_DIR/${report_num}-${id}.log"
+
+  # Prepare system prompt with placeholders resolved
+  local resolved_prompt="$BATCH_DIR/.resolved-prompt-${id}.md"
+  # Escape sed delimiter characters in variables to prevent substitution breakage
+  local esc_url esc_jd_file esc_report_num esc_date esc_id
+  esc_url="${url//\\/\\\\}"
+  esc_url="${esc_url//|/\\|}"
+  esc_jd_file="${jd_file//\\/\\\\}"
+  esc_jd_file="${esc_jd_file//|/\\|}"
+  esc_report_num="${report_num//|/\\|}"
+  esc_date="${date//|/\\|}"
+  esc_id="${id//|/\\|}"
+  sed \
+    -e "s|{{URL}}|${esc_url}|g" \
+    -e "s|{{JD_FILE}}|${esc_jd_file}|g" \
+    -e "s|{{REPORT_NUM}}|${esc_report_num}|g" \
+    -e "s|{{DATE}}|${esc_date}|g" \
+    -e "s|{{ID}}|${esc_id}|g" \
+    "$PROMPT_FILE" > "$resolved_prompt"
+
+  # Inject user-layer personalization into the temporary worker prompt.
+  # The resolved prompt is gitignored runtime state, so user profile data stays
+  # out of the system layer while batch scoring matches interactive scoring.
+  for context_file in "$PROJECT_DIR/modes/_profile.md" "$PROJECT_DIR/config/profile.yml"; do
+    if [[ -f "$context_file" ]]; then
+      {
+        printf '\n\n---\n\n'
+        printf '## Runtime personalization: %s\n\n' "${context_file#$PROJECT_DIR/}"
+        sed 's/^/    /' "$context_file"
+        printf '\n'
+      } >> "$resolved_prompt"
+    fi
+  done
+
+  # Launch claude -p worker.
+  # Model defaults to the Claude Max subscription default unless --model was
+  # passed. Building the command in an array keeps quoting safe regardless.
+  # --strict-mcp-config (with no --mcp-config) starts workers with no MCP
+  # servers: they only evaluate offers and need none. Without it each parallel
+  # worker inherits the parent session's MCP (e.g. Playwright) and they deadlock
+  # fighting over the single shared browser when --parallel > 1 (issue #506).
+  local -a claude_args=(-p --dangerously-skip-permissions --strict-mcp-config)
+  if [[ -n "$MODEL" ]]; then
+    claude_args+=(--model "$MODEL")
+  fi
+  claude_args+=(--append-system-prompt-file "$resolved_prompt" "$prompt")
+
+  local exit_code=0
+  local terminal_failure_recorded=false
+  local shim_retries=0
+  local max_shim_retries=4
+  while true; do
+    exit_code=0
+    claude "${claude_args[@]}" > "$log_file" 2>&1 || exit_code=$?
+
+    if [[ $exit_code -eq 0 ]]; then
+      break
+    fi
+
+    # Check for Claude Code npm shim swap (exit code 127 + command not found)
+    if [[ $exit_code -eq 127 ]] && grep -qE "(claude: command not found|claude:.*not found|cannot find.*claude)" "$log_file" && (( shim_retries < max_shim_retries )); then
+      shim_retries=$((shim_retries + 1))
+      echo "    ⏳ Claude command not found (shim swap detected). Retrying in 30s (attempt $shim_retries/$max_shim_retries)..."
+      sleep 30
+      continue
+    fi
+
+    if is_session_limit_log "$log_file"; then
+      mark_paused_rate_limit "$id" "$url" "$started_at" "$report_num" "$retries" "$log_file"
+      echo "    ⏸️  Session/rate limit reached; pausing batch without consuming retry budget."
+      terminal_failure_recorded=true
+      break
+    fi
+
+    if is_rate_limit_log "$log_file" && (( retries < MAX_RETRIES )); then
+      if (( RATE_LIMIT_SLEEP <= 0 )); then
+        mark_paused_rate_limit "$id" "$url" "$started_at" "$report_num" "$retries" "$log_file"
+        echo "    ⏸️  Rate limited and --rate-limit-sleep is 0; pausing batch without consuming retry budget."
+        terminal_failure_recorded=true
+        break
+      fi
+      retries=$((retries + 1))
+      local retry_completed_at
+      retry_completed_at=$(date -u +%Y-%m-%dT%H:%M:%SZ)
+      update_state "$id" "$url" "rate_limited" "$started_at" "$retry_completed_at" "$report_num" "-" "rate-limit; retrying after ${RATE_LIMIT_SLEEP}s" "$retries"
+      echo "    ⏳ Rate limited (attempt $retries/$MAX_RETRIES). Waiting ${RATE_LIMIT_SLEEP}s before retry..."
+      sleep "$RATE_LIMIT_SLEEP"
+      continue
+    fi
+
+    break
+  done
+
+  # Cleanup resolved prompt
+  rm -f "$resolved_prompt"
+
+  local completed_at
+  completed_at=$(date -u +%Y-%m-%dT%H:%M:%SZ)
+
+  if [[ $exit_code -eq 0 ]]; then
+    # Try to extract score from worker output
+    local score="-"
+    local score_match
+   score_match=$(sed -nE 's/.*"score":[[:space:]]*([0-9.]+).*/\1/p' "$log_file" 2>/dev/null | head -1 || true)
+    if [[ -n "$score_match" ]]; then
+      score="$score_match"
+    fi
+
+    # Check min-score gate
+    if is_decimal_number "$score" && awk -v min="$MIN_SCORE" 'BEGIN{exit !(min > 0)}'; then
+      if awk -v score="$score" -v min="$MIN_SCORE" 'BEGIN{exit !(score < min)}'; then
+        update_state "$id" "$url" "skipped" "$started_at" "$completed_at" "$report_num" "$score" "below-min-score" "$retries"
+        echo "    ⏭️  Skipped (score: $score < min-score: $MIN_SCORE)"
+        return 0
+      fi
+    fi
+
+    update_state "$id" "$url" "completed" "$started_at" "$completed_at" "$report_num" "$score" "-" "$retries"
+    echo "    ✅ Completed (score: $score, report: $report_num)"
+  elif [[ "$terminal_failure_recorded" == "false" ]]; then
+    if (( retries < MAX_RETRIES )); then
+      retries=$((retries + 1))
+    fi
+    local error_msg
+    error_msg=$(tail -5 "$log_file" 2>/dev/null | tr '\n' ' ' | cut -c1-200 || echo "Unknown error (exit code $exit_code)")
+    update_state "$id" "$url" "failed" "$started_at" "$completed_at" "$report_num" "-" "$error_msg" "$retries"
+    echo "    ❌ Failed (attempt $retries, exit code $exit_code)"
+  fi
+}
+
+# Merge tracker additions into applications.md
+merge_tracker() {
+  echo ""
+  echo "=== Merging tracker additions ==="
+  node "$PROJECT_DIR/merge-tracker.mjs"
+  echo ""
+  echo "=== Reconciling pipeline.md ==="
+  node "$PROJECT_DIR/reconcile-pipeline.mjs" || echo "⚠️  Pipeline reconcile had issues (see above)"
+  echo ""
+  echo "=== Verifying pipeline integrity ==="
+  node "$PROJECT_DIR/verify-pipeline.mjs" || echo "⚠️  Verification found issues (see above)"
+}
+
+# Print summary
+print_summary() {
+  echo ""
+  echo "=== Batch Summary ==="
+
+  if [[ ! -f "$STATE_FILE" ]]; then
+    echo "No state file found."
+    return
+  fi
+
+  local total=0 completed=0 skipped=0 failed=0 pending=0
+  local score_sum=0 score_count=0
+
+  while IFS=$'\t' read -r sid _ sstatus _ _ _ sscore _ _; do
+    [[ "$sid" == "id" ]] && continue
+    total=$((total + 1))
+    case "$sstatus" in
+      completed) completed=$((completed + 1))
+        if is_decimal_number "$sscore"; then
+          score_sum=$(awk -v sum="$score_sum" -v score="$sscore" 'BEGIN{print sum + score}' 2>/dev/null || echo "$score_sum")
+          score_count=$((score_count + 1))
+        fi
+        ;;
+      skipped) skipped=$((skipped + 1)) ;;
+      failed) failed=$((failed + 1)) ;;
+      *) pending=$((pending + 1)) ;;
+    esac
+  done < "$STATE_FILE"
+
+  echo "Total: $total | Completed: $completed | Skipped: $skipped | Failed: $failed | Pending: $pending"
+
+  if (( score_count > 0 )); then
+    local avg
+    avg=$(awk -v sum="$score_sum" -v count="$score_count" 'BEGIN{printf "%.1f", sum / count}' 2>/dev/null || echo "N/A")
+    echo "Average score: $avg/5 ($score_count scored)"
+  fi
+}
+
+print_status_table() {
+  if [[ ! -f "$STATE_FILE" ]]; then
+    echo "No state file found at $STATE_FILE"
+    return
+  fi
+
+  local total=0 completed=0 processing=0 failed=0 pending=0 skipped=0 rate_limited=0 paused_rate_limit=0
+  local score_sum=0 score_count=0
+
+  # Read first line to skip header
+  local header=true
+  while IFS=$'\t' read -r sid surl sstatus sstarted scompleted sreport sscore serror sretries || [[ -n "$sid" ]]; do
+    if [[ "$header" == "true" ]]; then
+      header=false
+      continue
+    fi
+    [[ -z "$sid" ]] && continue
+    sstatus="${sstatus%$'\r'}"
+    sscore="${sscore%$'\r'}"
+    serror="${serror%$'\r'}"
+    sreport="${sreport%$'\r'}"
+    total=$((total + 1))
+    case "$sstatus" in
+      completed)
+        completed=$((completed + 1))
+        if is_decimal_number "$sscore"; then
+          score_sum=$(awk -v sum="$score_sum" -v score="$sscore" 'BEGIN{print sum + score}' 2>/dev/null || echo "$score_sum")
+          score_count=$((score_count + 1))
+        fi
+        ;;
+      processing) processing=$((processing + 1)) ;;
+      failed) failed=$((failed + 1)) ;;
+      skipped) skipped=$((skipped + 1)) ;;
+      rate_limited) rate_limited=$((rate_limited + 1)) ;;
+      paused_rate_limit) paused_rate_limit=$((paused_rate_limit + 1)) ;;
+      *) pending=$((pending + 1)) ;;
+    esac
+  done < "$STATE_FILE"
+
+  echo "=== Batch Progress ==="
+  echo "Total: $total | Completed: $completed | Processing: $processing | Failed: $failed | Pending: $pending | Skipped: $skipped | Rate Limited: $rate_limited | Paused: $paused_rate_limit"
+  if (( score_count > 0 )); then
+    local avg
+    avg=$(awk -v sum="$score_sum" -v count="$score_count" 'BEGIN{printf "%.1f", sum / count}' 2>/dev/null || echo "N/A")
+    echo "Average score: $avg/5 ($score_count scored)"
+  fi
+  echo ""
+
+  # Format the per-job table:
+  # Columns: ID, Status, Report, Score, Target (URL or Error Message)
+  printf "%-4s | %-17s | %-6s | %-5s | %-40s\n" "ID" "Status" "Report" "Score" "URL / Error"
+  printf "%-4s+%-19s+%-8s+%-7s+%-42s\n" "----" "-------------------" "--------" "-------" "------------------------------------------"
+
+  header=true
+  while IFS=$'\t' read -r sid surl sstatus sstarted scompleted sreport sscore serror sretries || [[ -n "$sid" ]]; do
+    if [[ "$header" == "true" ]]; then
+      header=false
+      continue
+    fi
+    [[ -z "$sid" ]] && continue
+    sstatus="${sstatus%$'\r'}"
+    sscore="${sscore%$'\r'}"
+    serror="${serror%$'\r'}"
+    sreport="${sreport%$'\r'}"
+    local target="$surl"
+    if [[ "$sstatus" == "failed" && -n "$serror" && "$serror" != "-" ]]; then
+      target="Error: $serror"
+    fi
+    # Trim target to fit nicely (e.g. 50 chars)
+    if (( ${#target} > 50 )); then
+      target="${target:0:47}..."
+    fi
+    printf "%-4s | %-17s | %-6s | %-5s | %-50s\n" "$sid" "$sstatus" "$sreport" "$sscore" "$target"
+  done < "$STATE_FILE"
+}
+
+watch_status() {
+  local active_pid=""
+  if [[ -f "$LOCK_FILE" ]]; then
+    active_pid=$(cat "$LOCK_FILE" 2>/dev/null || true)
+  fi
+
+  if [[ -n "$active_pid" ]] && kill -0 "$active_pid" 2>/dev/null; then
+    echo "Watching batch-runner (PID $active_pid)... Press Ctrl+C to stop."
+    while kill -0 "$active_pid" 2>/dev/null; do
+      clear || printf "\033[c"
+      echo "=== Watching Batch Progress (PID $active_pid) ==="
+      print_status_table
+      sleep 2
+    done
+    echo ""
+    echo "=== Batch runner process (PID $active_pid) has finished ==="
+  else
+    echo "No active batch-runner detected."
+  fi
+
+  echo "Showing final status:"
+  print_status_table
+
+  # Chain verify-pipeline.mjs
+  if [[ -f "$PROJECT_DIR/verify-pipeline.mjs" ]]; then
+    echo ""
+    echo "=== Running pipeline verification ==="
+    node "$PROJECT_DIR/verify-pipeline.mjs" || echo "⚠️  Verification found issues"
+  fi
+}
+
+# Main
+main() {
+  if [[ "$STATUS_ONLY" == "true" ]]; then
+    check_status_prerequisites
+    print_status_table
+    exit 0
+  fi
+
+  if [[ "$WATCH_MODE" == "true" ]]; then
+    check_status_prerequisites
+    watch_status
+    exit 0
+  fi
+
+  check_prerequisites
+
+  if [[ "$DRY_RUN" == "false" ]]; then
+    acquire_lock
+    rm -f "$PAUSE_FILE"
+  fi
+
+  init_state
+
+  # Count input offers (skip header, ignore blank lines)
+  local total_input
+  total_input=$(tail -n +2 "$INPUT_FILE" | grep -c '[^[:space:]]' 2>/dev/null || true)
+  total_input="${total_input:-0}"
+
+  if (( total_input == 0 )); then
+    echo "No offers in $INPUT_FILE. Add offers first."
+    exit 0
+  fi
+
+  echo "=== career-ops batch runner ==="
+  if (( LIMIT > 0 )); then
+    echo "Parallel: $PARALLEL | Max retries: $MAX_RETRIES | Limit: $LIMIT"
+  else
+    echo "Parallel: $PARALLEL | Max retries: $MAX_RETRIES"
+  fi
+  echo "Input: $total_input offers"
+  echo ""
+
+  # Build list of offers to process
+  local -a pending_ids=()
+  local -a pending_urls=()
+  local -a pending_sources=()
+  local -a pending_notes=()
+
+  while IFS=$'\t' read -r id url source notes; do
+    [[ "$id" == "id" ]] && continue  # skip header
+    [[ -z "$id" || -z "$url" ]] && continue
+
+    # Guard against non-numeric id values
+    [[ "$id" =~ ^[0-9]+$ ]] || continue
+
+    # Skip if before start-from
+    if (( id < START_FROM )); then
+      continue
+    fi
+
+    local status
+    status=$(get_status "$id")
+
+    if [[ "$RESUME_PAUSED" == "true" ]]; then
+      if [[ "$status" != "paused_rate_limit" ]]; then
+        continue
+      fi
+    elif [[ "$RETRY_FAILED" == "true" ]]; then
+      # Only process failed offers
+      if [[ "$status" != "failed" ]]; then
+        continue
+      fi
+      # Check retry limit
+      local retries
+      retries=$(get_retries "$id")
+      if (( retries >= MAX_RETRIES )); then
+        echo "SKIP #$id: max retries ($MAX_RETRIES) reached"
+        continue
+      fi
+    else
+      # Skip terminal offers
+      if [[ "$status" == "completed" || "$status" == "skipped" ]]; then
+        continue
+      fi
+      # Paused rate-limit offers resume explicitly with --resume-paused.
+      if [[ "$status" == "paused_rate_limit" ]]; then
+        continue
+      fi
+      # Skip failed offers that hit retry limit (unless --retry-failed)
+      if [[ "$status" == "failed" ]]; then
+        local retries
+        retries=$(get_retries "$id")
+        if (( retries >= MAX_RETRIES )); then
+          echo "SKIP #$id: failed and max retries reached (use --retry-failed to force)"
+          continue
+        fi
+      fi
+    fi
+
+    if (( LIMIT > 0 )) && (( ${#pending_ids[@]} >= LIMIT )); then
+      break
+    fi
+
+    pending_ids+=("$id")
+    pending_urls+=("$url")
+    pending_sources+=("$source")
+    pending_notes+=("$notes")
+  done < "$INPUT_FILE"
+
+  local pending_count=${#pending_ids[@]}
+
+  if (( pending_count == 0 )); then
+    echo "No offers to process."
+    print_summary
+    exit 0
+  fi
+
+  echo "Pending: $pending_count offers"
+  echo ""
+
+  # Dry run: just list
+  if [[ "$DRY_RUN" == "true" ]]; then
+    echo "=== DRY RUN (no processing) ==="
+    for i in "${!pending_ids[@]}"; do
+      local status
+      status=$(get_status "${pending_ids[$i]}")
+      echo "  #${pending_ids[$i]}: ${pending_urls[$i]} [${pending_sources[$i]}] (status: $status)"
+    done
+    echo ""
+    echo "Would process $pending_count offers"
+    exit 0
+  fi
+
+  # Process offers
+  if (( PARALLEL <= 1 )); then
+    # Sequential processing
+    for i in "${!pending_ids[@]}"; do
+      process_offer "${pending_ids[$i]}" "${pending_urls[$i]}" "${pending_sources[$i]}" "${pending_notes[$i]}"
+      if [[ "$BATCH_PAUSED" == "true" || -f "$PAUSE_FILE" ]]; then
+        echo "=== Batch paused: session/rate limit reached. Resume later with --resume-paused. ==="
+        break
+      fi
+    done
+  else
+    # Parallel processing with job control
+    local running=0
+    local -a pids=()
+    local -a pid_ids=()
+
+    for i in "${!pending_ids[@]}"; do
+      if [[ "$BATCH_PAUSED" == "true" || -f "$PAUSE_FILE" ]]; then
+        echo "=== Batch paused: session/rate limit reached. Waiting for running workers, not scheduling new offers. ==="
+        break
+      fi
+
+      # Wait if we're at parallel limit
+      while (( running >= PARALLEL )); do
+        # Wait for any child to finish
+        for j in "${!pids[@]}"; do
+          if ! kill -0 "${pids[$j]}" 2>/dev/null; then
+            wait "${pids[$j]}" 2>/dev/null || true
+            unset 'pids[j]'
+            unset 'pid_ids[j]'
+            running=$((running - 1))
+          fi
+        done
+        # Compact arrays
+        pids=("${pids[@]}")
+        pid_ids=("${pid_ids[@]}")
+        if [[ "$BATCH_PAUSED" == "true" || -f "$PAUSE_FILE" ]]; then
+          echo "=== Batch paused: session/rate limit reached. Waiting for running workers, not scheduling new offers. ==="
+          break
+        fi
+        sleep 1
+      done
+
+      if [[ "$BATCH_PAUSED" == "true" || -f "$PAUSE_FILE" ]]; then
+        break
+      fi
+
+      # Launch worker in background
+      process_offer "${pending_ids[$i]}" "${pending_urls[$i]}" "${pending_sources[$i]}" "${pending_notes[$i]}" &
+      pids+=($!)
+      pid_ids+=("${pending_ids[$i]}")
+      running=$((running + 1))
+    done
+
+    # Wait for remaining workers
+    for pid in "${pids[@]}"; do
+      wait "$pid" 2>/dev/null || true
+    done
+  fi
+
+  # Merge tracker additions
+  merge_tracker
+
+  # Print summary
+  print_summary
+
+  exit 0
+}
+
+main "$@"
+
@@ -0,0 +1,11 @@
+id	url	status	started_at	completed_at	report_num	score	error	retries
+1	https://getmatch.ru/vacancies/34414-starshii-frontend-razrabotchik?s=offers	completed	2026-07-08T04:20:29Z	2026-07-08T04:23:49Z	066	-	-	0
+3	https://hh.ru/vacancy/133613060?query=frontend&hhtmFrom=vacancy_search_list	completed	2026-07-08T04:20:29Z	2026-07-08T04:22:50Z	067	1.0	-	0
+2	https://hirehi.ru/development/frontend-razrabotchik-55966	completed	2026-07-08T04:20:29Z	2026-07-08T04:23:38Z	068	1.2	-	0
+4	https://hh.ru/vacancy/133056126?query=frontend&hhtmFrom=vacancy_search_list	completed	2026-07-08T04:22:52Z	2026-07-08T04:25:32Z	069	1.2	-	0
+5	https://getmatch.ru/vacancies/35037-frontend-software-engineer-bootcamp?s=offers	completed	2026-07-08T04:23:40Z	2026-07-08T04:26:23Z	070	1.0	-	0
+6	https://career.habr.com/vacancies/1000167021	completed	2026-07-08T04:23:51Z	2026-07-08T04:26:50Z	071	1.8	-	0
+7	https://hh.ru/vacancy/119867578?query=frontend&hhtmFrom=vacancy_search_list	completed	2026-07-08T04:25:34Z	2026-07-08T04:29:23Z	072	1.7	-	0
+8	https://t.me/s/runello_rus_javascript/1443	completed	2026-07-08T04:26:24Z	2026-07-08T04:27:23Z	073	-	-	0
+9	https://t.me/s/runello_rus_javascript/1445	completed	2026-07-08T04:26:51Z	2026-07-08T04:29:25Z	074	-	-	0
+10	https://t.me/s/runello_rus_javascript/1448	completed	2026-07-08T04:27:24Z	2026-07-08T04:31:35Z	075	3.0	-	0
@@ -0,0 +1,7 @@
+**VK — Старший Frontend-разработчик (React): Score 1.0/5, Skip.**
+
+Hallazgo clave: al hacer WebFetch de la URL, la fuente indica explícitamente que **la vacante está archivada** (VK ya no busca candidatos) — el JD file estaba vacío, así que este dato solo salió al intentar el fallback. Además de eso, dos bloqueadores estructurales independientes del estado de la vacante:
+- Stack **React** puro (hooks, context, state management) — sin coincidencia con la especialización Vue 3/Flutter del perfil.
+- Nivel **Senior 5+ años** vs ~1.5-2 años de experiencia real de la candidata.
+
+Comp era en realidad excelente (300k+₽, por encima del rango objetivo), pero irrelevante dado lo anterior. Report guardado en `reports/066-vk-2026-07-08.md`, línea de tracker en `batch/tracker-additions/066-vk.tsv` (próximo número: 16), PDF no generado (score bajo el umbral de 3.0).
@@ -0,0 +1,32 @@
+## Resultado: oferta duplicada, no se generó nueva evaluación
+
+Esta URL (`https://hh.ru/vacancy/133613060?query=frontend&hhtmFrom=vacancy_search_list`) ya fue evaluada ayer como **report 053** — es la vacante exacta de **Студия Перцы** (Ярославль), programa de trainee para principiantes:
+
+- Стек: React + Sass (verstka/animación para juegos/promo-sitios), no Vue
+- Nivel: sin experiencia requerida (programa para estudiantes/graduados) — la candidata ya es Middle con experiencia comercial
+- Formato: **solo presencial en Ярославль**, sin remoto (bloqueador duro según `modes/_profile.md`: candidata en Тула, sin reubicación)
+- Salario: 40 000–50 000 ₽/mes (etapa de stажировка: 20 000 ₽) — muy por debajo del mínimo objetivo (100 000–120 000+ ₽)
+- **Score:** 1.0/5 · **Legitimacy:** High Confidence (empresa real, activa desde 2006, reseñas verificadas)
+
+Confirmé que:
+- El TSV de tracker (`batch/tracker-additions/053-studiya-percy.tsv`) ya existe para esta misma empresa+rol, pendiente de `merge-tracker.mjs`.
+- Report `053-studiya-percy-2026-07-07.md` cubre exactamente el mismo JD/URL.
+
+Por la regla de integridad del pipeline (nunca duplicar entradas de company+role ya evaluadas), **no generé el report 067 ni un nuevo tracker line** — sería una duplicación exacta del mismo posting evaluado ayer.
+
+```json
+{
+  "status": "skipped_duplicate",
+  "id": "3",
+  "report_num": "067",
+  "company": "Студия Перцы",
+  "role": "Frontend-разработчик",
+  "score": 1.0,
+  "legitimacy": "High Confidence",
+  "pdf": null,
+  "report": "reports/053-studiya-percy-2026-07-07.md",
+  "error": "Duplicate of existing report 053 (same hh.ru vacancy URL, already evaluated 2026-07-07). No new report or tracker line created per pipeline integrity rule."
+}
+```
+
+Cuando ejecutes `node merge-tracker.mjs`, la entrada #053 (score 1.0/5, no aplicar) se fusionará en `data/applications.md` normalmente — no requiere ninguna acción adicional de esta corrida.
@@ -0,0 +1,18 @@
+Evaluación completa. Este es un **SKIP** claro: la posting está marcada por la fuente como archivada/cerrada, el formato es híbrido en Moscú (bloqueo duro de ubicación — la candidata no se muda de Тула), el stack es React/Redux (no Vue.js, su arquetipo primario), y pide iOS nativo en Swift (solo tiene Flutter cross-platform).
+
+```json
+{
+  "status": "completed",
+  "id": "2",
+  "report_num": "068",
+  "company": "Сбер (СберТех)",
+  "role": "Frontend-разработчик (Middle)",
+  "score": 1.2,
+  "legitimacy": "Suspicious",
+  "pdf": null,
+  "report": "reports/068-sber-frontend-2026-07-08.md",
+  "error": null
+}
+```
+
+Resumen: score 1.2/5, PDF no generado (por debajo del umbral 3.0), report guardado, línea de tracker escrita en `batch/tracker-additions/2.tsv` (pendiente de `node merge-tracker.mjs`).
@@ -0,0 +1,18 @@
+Tracker line and report are done. Final JSON summary for the orchestrator:
+
+```json
+{
+  "status": "completed",
+  "id": "4",
+  "report_num": "069",
+  "company": "Т-Банк",
+  "role": "Frontend Developer (React)",
+  "score": 1.2,
+  "legitimacy": "High Confidence",
+  "pdf": null,
+  "report": "reports/069-t-bank-2026-07-08.md",
+  "error": null
+}
+```
+
+**Resumen:** Т-Банк, Frontend Developer (React) en Kazan — score 1.2/5, SKIP claro. Tres bloqueadores duros: stack React (no Vue, el arquetipo principal de la candidata), seniority requerida de 3-6 años (vs. ~1.5-2 años reales), y ubicación en Kazan que exige reubicación (política explícita de no-reubicación). No se generó PDF (por debajo del umbral 3.0). Report guardado en `reports/069-t-bank-2026-07-08.md`, línea de tracker en `batch/tracker-additions/4.tsv` (pendiente de `node merge-tracker.mjs`).
@@ -0,0 +1,18 @@
+Now the final JSON summary for the orchestrator:
+
+```json
+{
+  "status": "completed",
+  "id": "5",
+  "report_num": "070",
+  "company": "Exness",
+  "role": "Frontend Software Engineer (Bootcamp)",
+  "score": 1.0,
+  "legitimacy": "High Confidence",
+  "pdf": null,
+  "report": "reports/070-exness-2026-07-08.md",
+  "error": null
+}
+```
+
+**Resumen:** Score 1.0/5 — SKIP. Esta oferta (Exness, Limassol/Chipre, oficina) es un **duplicado exacto** del reporte 055 (2026-07-07, mismo ID de vacante 35037), confirmado en `data/scan-history.tsv`. Bloqueadores duros sin cambios: puesto presencial en Chipre que exige residencia previa allí (política de no reubicación de la candidata) + stack React (la candidata trabaja en Vue 3). No se generó PDF (score bajo el umbral 3.0). Report y línea de tracker guardados; falta correr `node merge-tracker.mjs` para fusionar el batch completo.
@@ -0,0 +1,19 @@
+Report saved. Score is well below the 3.0 PDF auto-generation threshold anyway (and it's a duplicate), so no PDF was generated, and no tracker line was written for batch ID 6 to prevent a duplicate row in `applications.md`.
+
+```json
+{
+  "status": "completed",
+  "id": "6",
+  "report_num": "071",
+  "company": "РТК-ЦОД",
+  "role": "Ведущий фронтенд-разработчик (React)",
+  "score": 1.8,
+  "legitimacy": "High Confidence",
+  "pdf": null,
+  "report": "reports/071-rtk-tsod-2026-07-08.md",
+  "error": null,
+  "note": "Duplicate of report 056 (same URL/JD, evaluated 2026-07-07). Tracker line intentionally skipped to avoid a duplicate applications.md row — company+role already covered by batch/tracker-additions/056-rtk-tsod.tsv."
+}
+```
+
+**Resumen:** РТК-ЦОД — Ведущий фронтенд-разработчик (React), 1.8/5, no recomendado. Es un duplicado exacto de la oferta ya evaluada ayer (reporte 056) — mismo URL, misma JD. Bloqueadores: stack React/Next.js/GraphQL sin ningún proyecto React en el CV (todo el perfil es Vue 3 + Flutter), y nivel Senior/Lead vs. Middle del candidato. No se generó PDF ni línea de tracker nueva para evitar duplicar la entrada de РТК-ЦОД que ya está pendiente de merge desde el reporte 056.
@@ -0,0 +1,16 @@
+```json
+{
+  "status": "completed",
+  "id": "7",
+  "report_num": "072",
+  "company": "Т-Банк",
+  "role": "Frontend-разработчик React (Junior)",
+  "score": 1.7,
+  "legitimacy": "Proceed with Caution",
+  "pdf": null,
+  "report": "reports/072-t-bank-react-junior-perm-2026-07-08.md",
+  "error": null
+}
+```
+
+Evaluated Т-Банк's Junior React frontend role (Perm, remote-friendly) — scored 1.7/5, recommendation SKIP. Main blockers: role requires React (candidate's primary stack is Vue.js, no commercial React experience) and the Junior scope/comp sits below her actual level of ownership (~100k+ ₽/mes minimum vs. Junior market rates of ~55-80k). No PDF generated (below 3.0 threshold). Report saved to `reports/072-t-bank-react-junior-perm-2026-07-08.md`, tracker line queued at `batch/tracker-additions/7.tsv` — run `node merge-tracker.mjs` when the batch finishes.
@@ -0,0 +1,17 @@
+No report, PDF, or tracker line was created for this ID since there is no actual job content to evaluate.
+
+```json
+{
+  "status": "failed",
+  "id": "8",
+  "report_num": "073",
+  "company": "unknown",
+  "role": "unknown",
+  "score": null,
+  "pdf": null,
+  "report": null,
+  "error": "JD file /tmp/batch-jd-8.okc7U0 is empty (0 bytes), and the fallback URL https://t.me/s/runello_rus_javascript/1443 does not contain a job posting — message #1443 in that channel is a Runello bot promotional/service message ('upload your resume to @RunelloBot'), not an actual JD. No evaluable content exists for this offer."
+}
+```
+
+Report #073 was not generated — this batch ID has no real job description to score. Recommend removing this URL from the batch input or re-scraping the correct message ID for that channel entry.
@@ -0,0 +1 @@
+**Resumen:** Esta oferta (`t.me/s/runello_rus_javascript/1445`) es un duplicado exacto del post ya evaluado como reporte 059 el 2026-07-07 — mismo empleador anónimo, mismo stack (React + Python/LLM/LangChain/RAG), mismo bot de Runello como único canal de aplicación. Score 1.0/5, legitimidad **Suspicious** (empleador no revelado + aplicación forzada vía bot de terceros), sin match real con el perfil Vue/Flutter de la candidata. No se generó PDF (por debajo del umbral 3.0). Reporte guardado en `reports/074-unknown-employer-runello-2026-07-08.md` y línea de tracker en `batch/tracker-additions/9.tsv`. Recomendación: **no aplicar**.
@@ -0,0 +1,25 @@
+**Duplicate detected — no new report/PDF created.**
+
+URL `https://t.me/s/runello_rus_javascript/1448` (batch ID 10) is identical to a posting already fully evaluated yesterday as **report 060** (`reports/060-konsultantplyus-2026-07-07.md`, ЗАО КонсультантПлюс — Веб разработчик Vue/Js, score 3.0/5, Proceed with Caution). The JD file was empty and WebFetch returned the same thin Telegram/Runello preview (only title + company blurb — no requirements, salary, or location), so there is no new information that would change that evaluation.
+
+Per the pipeline-integrity rule ("never create duplicate entries for the same company+role"), I did not generate report 075 or a new PDF. Instead I found that report 060's tracker line had never been written (a pre-existing gap — its score never made it into `data/applications.md`), so I closed that gap:
+
+- Wrote `batch/tracker-additions/10.tsv` pointing to the existing report 060 (score 3.0/5, ❌ PDF — the PDF was never actually compiled despite the report claiming it hit the 3.0 threshold; only the HTML exists in `output/`).
+- Run `node merge-tracker.mjs` when convenient to fold this into `data/applications.md`.
+
+```json
+{
+  "status": "completed",
+  "id": "10",
+  "report_num": "075",
+  "company": "ЗАО КонсультантПлюс",
+  "role": "Веб разработчик Vue/Js",
+  "score": 3.0,
+  "legitimacy": "Proceed with Caution",
+  "pdf": null,
+  "report": "reports/060-konsultantplyus-2026-07-07.md",
+  "error": "duplicate_of_report_060: identical URL/JD already evaluated 2026-07-07; no new report or PDF generated; wrote missing tracker line for existing report 060 instead"
+}
+```
+
+If the scanner is re-feeding already-processed URLs into the batch inbox, it's worth checking `data/scan-history.tsv` dedup logic for why this one resurfaced.
@@ -0,0 +1 @@
+001	2026-07-06	Яндекс (Лавка)	Разработчик фронтенда (Vue/React) в Лавку	Evaluated	1.0/5	❌	[001](reports/001-yandex-lavka-2026-07-06.md)	Сильный стек-мэтч (Vue3+тесты+legacy+ритейл), но офис/гибрид в Москве/СПб без удалёнки — жёсткий локационный блокер
@@ -0,0 +1 @@
+002	2026-07-06	VK	Frontend-разработчик (MAX)	Evaluated	2.3/5	❌	[002](reports/002-vk-max-2026-07-06.md)	Удалёнка доступна и уровень Middle совпадает, но ядро роли — React-экосистема, которой нет в CV (стек кандидата — Vue)
@@ -0,0 +1 @@
+003	2026-07-06	VK	Старший Frontend-разработчик сообщества (ВКонтакте)	Evaluated	1.8/5	❌	[003](reports/003-vk-vkontakte-senior-2026-07-06.md)	Двойной разрыв: требуется Senior/4+ лет и React, у кандидата Middle/~2.5 года и стек Vue; не рекомендуется откликаться
@@ -0,0 +1 @@
+004	2026-07-06	Ozon	Frontend-разработчик (ERP и учётные системы)	Evaluated	4.2/5	✅	[004](reports/004-ozon-2026-07-06.md)	Лучший фит пакета: Middle Vue3/TS, удалённо, formal "3+ года JS" порог выше факта (~2 года)
@@ -0,0 +1 @@
+005	2026-07-06	Ozon	Стажёр Frontend-разработчик (Маркетинговая платформа)	Evaluated	2.5/5	❌	[005](reports/005-ozon-2026-07-06.md)	Уровень "стажёр" ниже Middle-профиля кандидата
@@ -0,0 +1 @@
+006	2026-07-06	Wildberries	Frontend-разработчик (Платформа)	Evaluated	2.8/5	❌	[006](reports/006-wildberries-2026-07-06.md)	Стек React, не Vue — adjacent архетип, не primary
@@ -0,0 +1 @@
+007	2026-07-06	Wildberries (RWB)	Senior Frontend developer (Angular) в WB Job	Evaluated	1.7/5	❌	[007](reports/007-wildberries-2026-07-06.md)	Удалённо, но нет Angular и стаж вдвое короче требуемых 4 лет — строго не рекомендуется
@@ -0,0 +1 @@
+008	2026-07-06	Т-Банк	Frontend-разработчик — React	Evaluated	2.0/5	❌	[008](reports/008-tbank-react-2026-07-06.md)	Удалённо/гибко, но нет коммерческого React (обязательно от 4 лет) — строго не рекомендуется
@@ -0,0 +1 @@
+009	2026-07-06	Т-Банк	Frontend-разработчик — Angular	Evaluated	2.3/5	❌	[009](reports/009-tbank-angular-2026-07-06.md)	Удалённо/гибко, Middle/Senior, но нет Angular (обязательно от 2 лет) — строго не рекомендуется
@@ -0,0 +1 @@
+010	2026-07-06	МТС	Senior Frontend разработчик [NBA.Банк]	Evaluated	1.6/5	❌	[010](reports/010-mts-senior-frontend-nba-bank-2026-07-06.md)	Сильное совпадение по Vue3/TS, но гибрид в Москве нарушает жёсткое правило локации (Тула, без переезда) — балл тонет до 1.6
@@ -0,0 +1 @@
+011	2026-07-06	МТС	Intern Frontend dev [Governance]	Evaluated	2.0/5	❌	[011](reports/011-mts-intern-frontend-governance-2026-07-06.md)	Полностью удалённо (ок для локации), но стек React (не Vue) и уровень "стажёр без опыта" не подходят Middle-кандидату
@@ -0,0 +1 @@
+012	2026-07-06	МТС	Middle Frontend Developer (Vue) [GARNET]	Evaluated	1.9/5	❌	[012](reports/012-mts-middle-frontend-vue-garnet-2026-07-06.md)	Лучшее стековое совпадение в пакете (Vue3+Composition API+TS), но тот же блокер: гибрид в Москве, не Тула
@@ -0,0 +1 @@
+013	2026-07-06	МТС	Senior Frontend (ON Медиа)	Evaluated	1.7/5	❌	[013](reports/013-mts-on-media-2026-07-06.md)	Стек Angular не совпадает с primary-архетипом (Vue.js)
@@ -0,0 +1 @@
+014	2026-07-06	МТС	Senior Frontend разработчик	Evaluated	1.6/5	❌	[014](reports/014-mts-frontend-razrabotchik-2026-07-06.md)	Стек Vue совпадает, но гибрид/офис в Москве — жёсткий локационный блокер
@@ -0,0 +1 @@
+015	2026-07-06	Kaspersky	JavaScript Developer (KES)	Evaluated	3.2/5	✅	[015](reports/015-kaspersky-kes-2026-07-06.md)	React+Node enterprise-роль, удалённо по всей России, комп сильный, стек — умеренный разрыв
@@ -0,0 +1 @@
+016	2026-07-07	КСК ЭЛКОМ	Разработчик Flutter (Dart)	Evaluated	1.0/5	❌	[016](reports/016-ksk-elkom-2026-07-07.md)	Офис в СПб без удалёнки — блокер по location-политике
@@ -0,0 +1 @@
+017	2026-07-07	АО "ЦФТ"	Frontend-разработчик Vue	Evaluated	2.1/5	❌	[017](reports/017-tsft-2026-07-07.md)	Сильное совпадение по Vue3/TS/Pinia, но стажевый пробел (3-6 лет vs ~1.7), формат работы не подтверждён (Новосибирск), зарплата не раскрыта
@@ -0,0 +1 @@
+018	2026-07-07	Яндекс	Fullstack-разработчик в команду frontend-инфраструктуры	Evaluated	1.0/5	❌	[018](reports/018-yandex-frontend-infra-2026-07-07.md)	Екатеринбург/Москва, гибрид/офис без удалёнки + роль platform-tooling, не UI
@@ -0,0 +1 @@
+019	2026-07-07	Яндекс	Разработчик фронтенда в команду Яндекс Кит	Evaluated	1.0/5	❌	[019](reports/019-yandex-kit-2026-07-07.md)	СПб/Москва без удалёнки + стек React вместо Vue
@@ -0,0 +1 @@
+020	2026-07-07	Яндекс	Мобильный разработчик (Android/iOS) Payment SDK в Финтех	Evaluated	1.0/5	❌	[020](reports/020-yandex-payment-sdk-2026-07-07.md)	СПб/Москва без удалёнки + нативный Kotlin/Swift вместо Flutter
@@ -0,0 +1 @@
+021	2026-07-07	Яндекс	Стажёр в команду разработки AI-инфраструктуры фронтенда Яндекс 360	Evaluated	1.0/5	❌	[021](reports/021-yandex-360-intern-2026-07-07.md)	Москва, только офис, стажировка на React — не подходит по уровню и локации
@@ -0,0 +1 @@
+022	2026-07-07	Яндекс	Фронтенд-разработчик в еком-сценарии Поиска	Evaluated	1.0/5	❌	[022](reports/022-yandex-ecom-search-2026-07-07.md)	Хороший содержательный фит, но СПб/Москва без удалёнки — блокер
@@ -0,0 +1 @@
+023	2026-07-07	Яндекс	Разработчик фронтенда в Яндекс 360	Evaluated	1.0/5	❌	[023](reports/023-yandex-360-frontend-2026-07-07.md)	СПб/Москва без удалёнки + React вместо Vue
@@ -0,0 +1 @@
+024	2026-07-07	Бизнес Технологии	Senior/Tech Lead frontend-разработчик	Evaluated	1.5/5	❌	[024](reports/024-biznes-tehnologii-2026-07-07.md)	Отличная компенсация (300-350к на руки), но требование 10 лет опыта vs 1.7 года у кандидата - критический разрыв
@@ -0,0 +1 @@
+025	2026-07-07	ИП Каржева М.К.	Frontend-разработчик	Evaluated	2.6/5	❌	[025](reports/025-karzheva-2026-07-07.md)	WebFetch дважды не смог извлечь JD; компенсация (~113-130к на руки расч.) в целевом диапазоне, но легитимность/детали не проверены
@@ -0,0 +1 @@
+026	2026-07-07	Dev Company 23	Frontend разработчик Senior	Evaluated	2.4/5	❌	[026](reports/026-dev-company-23-2026-07-07.md)	Сильнейшее совпадение стека (Vue3/TS/Pinia) и отличная компенсация (от 270к gross), но заявлено 4-6 лет опыта vs 1.7 года у кандидата
@@ -0,0 +1 @@
+027	2026-07-07	Sakura Games	Fullstack Developer (Node.js/TS + React/Next.js)	Evaluated	1.9/5	❌	[027](reports/027-sakura-games-2026-07-07.md)	Стек React/Next.js (не Vue) - нет production опыта; заявлено 6+ лет vs 1.7 года; зарплата не раскрыта; контракт проектный 3-6 мес
@@ -0,0 +1 @@
+028	2026-07-07	Selecty	Senior Frontend Developer (Vue.js/Nuxt.js)	Evaluated	2.0/5	❌	[028](reports/028-selecty-2026-07-07.md)	Удалёнка ок, но 5 лет опыта + английский B2+ mandatory — два hard-блокера
@@ -0,0 +1 @@
+029	2026-07-07	Bnovo	Senior Frontend разработчик	Evaluated	2.5/5	❌	[029](reports/029-bnovo-2026-07-07.md)	Удалёнка подтверждена, отличная репутация, но 4+ года опыта + tech-lead трек выше уровня
@@ -0,0 +1 @@
+030	2026-07-07	ООО Бизнес Технологии	Middle frontend-разработчик	Evaluated	4.2/5	✅	[030](reports/030-biznes-tehnologii-middle-2026-07-07.md)	Гибкий формат (вкл. удалёнку), 150-200к на руки, стек совпадает — рекомендуется откликнуться
@@ -0,0 +1 @@
+032	2026-07-07	ИП Шевченко М.Ю.	Flutter-разработчик	Evaluated	2.7/5	❌	[032](reports/032-ip-shevchenko-2026-07-07.md)	Проектный ГПХ-контракт, заявлен стаж 3-6 лет vs фактический ~1.5-2 года, агрессивный дедлайн (10% готовности за 8 недель)
@@ -0,0 +1 @@
+033	2026-07-07	МЕДИА ПОТОК	Fullstack-разработчик мобильных приложений	Evaluated	4.1/5	✅	[033](reports/033-media-potok-2026-07-07.md)	Стаж 1-3 года и стек (Flutter+Node.js) совпадают почти идеально, полностью гибкий формат, вилка 100-200к на руки
@@ -0,0 +1 @@
+034	2026-07-07	Playneta	Senior Flutter Developer	Evaluated	1.8/5	❌	[034](reports/034-playneta-2026-07-07.md)	Senior-геймдев роль (5+ лет, рендеринг/анимация 60-120fps) vs фактический опыт ~1.5-2 года без игрового бэкграунда; 2 дублирующих объявления замечены
@@ -0,0 +1 @@
+035	2026-07-07	Noroots	Fullstack-разработчик	Evaluated	N/A	❌	[035](reports/035-noroots-2026-07-07.md)	Вакансия недоступна для просмотра (403/ограничен доступ) — оценка невозможна
@@ -0,0 +1 @@
+036	2026-07-07	МТС	Senior Frontend developer [МТС Маркетолог]	Evaluated	1.8/5	❌	[036](reports/036-mts-marketolog-2026-07-07.md)	React vs Vue stack mismatch + Senior 6+ лет vs ~2 года опыта — не рекомендуется
@@ -0,0 +1 @@
+037	2026-07-07	МТС	Middle Flutter Developer [GOODOK]	Evaluated	3.6/5	✅	[037](reports/037-mts-goodok-2026-07-07.md)	Лучший фит партии: BLoC/iOS-портирование совпадают, remote-опция, но ниже порога 4.0
@@ -0,0 +1 @@
+038	2026-07-07	МТС	Fullstack developer React/Node.JS [MWS Tables]	Evaluated	1.9/5	❌	[038](reports/038-mts-mwstables-2026-07-07.md)	React + тяжёлый DevOps (K8s/Helm) далеко от Vue + ознакомительного Node — не рекомендуется
@@ -0,0 +1 @@
+039	2026-07-07	МТС	Стажер Fullstack QA [MWS Octapi]	Evaluated	1.0/5	❌	[039](reports/039-mts-octapi-2026-07-07.md)	Hard location blocker (Москва гибрид, без remote) + программа для студентов, не для выпускников
@@ -0,0 +1 @@
+040	2026-07-07	МТС	Junior Fullstack-аналитик [CICADA8]	Evaluated	1.0/5	❌	[040](reports/040-mts-cicada8-2026-07-07.md)	Hard location blocker (офис Москва, без remote) + роль аналитика не входит в архетипы кандидата
@@ -0,0 +1 @@
+051	2026-07-07	VK	Старший Frontend-разработчик (React, мессенджер)	Evaluated	1.5/5	❌	[051](reports/051-vk-messenger-frontend-2026-07-07.md)	Архивная вакансия, стек React (не Vue) и требование 5+ лет — тройной hard blocker, не рекомендуется откликаться
@@ -0,0 +1 @@
+052	2026-07-07	Сбер	Frontend Developer	Evaluated	1.0/5	❌	[052](reports/052-sber-2026-07-07.md)	Стек React/Redux/Swift/AI-моделей не совпадает с Vue/Flutter кандидата; роль в Москве требует переезда (hard blocker); листинг архивирован по данным hirehi.ru
@@ -0,0 +1 @@
+053	2026-07-07	Студия Перцы	Frontend-разработчик	Evaluated	1.0/5	❌	[053](reports/053-studiya-percy-2026-07-07.md)	Стажёрская программа для новичков, офис только в Ярославле без удалёнки, ЗП 40-50к — не совпадает по формату работы и уровню
@@ -0,0 +1 @@
+054	2026-07-07	Т-Банк	Frontend Developer (React)	Evaluated	1.0/5	❌	[054](reports/054-t-bank-2026-07-07.md)	Hard location blocker — гибрид офис Казань, без переезда; стек React vs Vue у кандидата
@@ -0,0 +1 @@
+055	2026-07-07	Exness	Frontend Software Engineer (Bootcamp)	Evaluated	1.0/5	❌	[055](reports/055-exness-2026-07-07.md)	Офис в Лимасоле (Кипр), требует уже жить на Кипре, стек React не Vue — hard location blocker, не рекомендуется
@@ -0,0 +1 @@
+056	2026-07-07	РТК-ЦОД	Ведущий фронтенд-разработчик (React)	Evaluated	1.8/5	❌	[056](reports/056-rtk-tsod-2026-07-07.md)	Стек React/Next.js/GraphQL и уровень Lead не совпадают с профилем (Vue 3, Middle) — не рекомендуется откликаться
@@ -0,0 +1 @@
+057	2026-07-07	Т-Банк	Frontend-разработчик React (junior)	Evaluated	2.4/5	❌	[057](reports/057-t-bank-2026-07-07.md)	Junior-грейд vs Middle кандидата, стек React не совпадает с Vue, зарплата не раскрыта — не рекомендуется
@@ -0,0 +1 @@
+059	2026-07-07	Unknown (Runello aggregator)	Fullstack React Developer	Evaluated	1.0/5	❌	[059](reports/059-unknown-employer-runello-2026-07-07.md)	Работодатель не назван, отклик только через сторонний бот; стек React+Python/LLM/RAG не совпадает с профилем (Vue+Flutter) — Suspicious legitimacy, не рекомендуется
@@ -0,0 +1 @@
+061	2026-07-07	TOOLUM	Fullstack разработчик Next.js	Evaluated	2.3/5	❌	[061](reports/061-toolum-2026-07-07.md)	Next.js/React hard req vs Vue-only CV; JD terse (no salary/location/format); TOOLUM no public footprint
@@ -0,0 +1 @@
+16	2026-07-08	VK	Старший Frontend-разработчик (React)	Evaluated	1.0/5	❌	[066](reports/066-vk-2026-07-08.md)	Вакансия архивирована (компания не ищет кандидатов) + стек React вместо Vue + Senior 5+ лет vs ~1.5 года опыта — SKIP
@@ -0,0 +1 @@
+075	2026-07-08	СПБ ИАЦ	Программист Front-end	Evaluated	1.0/5	❌	[075](reports/075-spb-iac-2026-07-08.md)	Vue 3/TS отлично совпадает, но офис только в Санкт-Петербурге без удалёнки — блокер по локационной политике
@@ -0,0 +1 @@
+076	2026-07-08	Bellerage	Стажер Frontend-разработчик	Evaluated	1.0/5	❌	[076](reports/076-bellerage-2026-07-08.md)	React (не Vue) + стажёрский уровень + офис в Москве без удалёнки — блокер по локации
@@ -0,0 +1 @@
+077	2026-07-08	Bioniq	Младший фронтенд-разработчик	Evaluated	1.0/5	❌	[077](reports/077-bioniq-2026-07-08.md)	Angular (не Vue) + вёрстка лендингов + офис в Москве — блокер по стеку и локации
@@ -0,0 +1 @@
+078	2026-07-08	А7-Технологии	Flutter-разработчик	Evaluated	1.0/5	❌	[078](reports/078-a7-tehnologii-2026-07-08.md)	Хорошее совпадение по Flutter/BLoC, но офис/гибрид в Москве без полной удалёнки — блокер по локации
@@ -0,0 +1 @@
+079	2026-07-08	DataLouna	Backend Engineer (Node.js, TypeScript, Data, Real-time)	Evaluated	1.0/5	❌	[079](reports/079-datalouna-2026-07-08.md)	Remote ок, но senior data/backend роль (ClickHouse, K8s, парсеры) без frontend — глубокое несовпадение по функции и опыту (5+ лет)
@@ -0,0 +1 @@
+080	2026-07-08	Selfwork	Full-stack engineer	Evaluated	1.5/5	❌	[080](reports/080-selfwork-2026-07-08.md)	Remote, TS/Node пересекаются, но React (не Vue) + жёсткое требование 5+ лет senior fullstack опыта не в аутсорсе
@@ -0,0 +1 @@
+081	2026-07-08	Positive Technologies	Fullstack QA Engineer (RA VPN)	Evaluated	1.0/5	❌	[081](reports/081-positive-technologies-2026-07-08.md)	Remote ок, но это QA-роль (тестирование VPN/NGFW), а не разработка — функциональное несовпадение
@@ -0,0 +1 @@
+082	2026-07-08	Air Production	Frontend-разработчик (Vue.js)	Evaluated	4.0/5	✅	[082](reports/082-air-production-2026-07-08.md)	Лучшее совпадение сессии: Vue.js, удалённо, опыт 1-3 года и ЗП 90-130к точно в цель; toolchain-разрыв (Vuex/Webpack/Gulp vs Pinia/Vite) закрываем
@@ -0,0 +1 @@
+083	2026-07-08	Nexters	Senior Frontend Developer (Hero Wars)	Evaluated	1.0/5	❌	[083](reports/083-nexters-2026-07-08.md)	Remote ок, но gamedev-стек (игровые движки Cocos/Phaser/Pixi) + Senior 3+ года — не веб-Vue разработка
@@ -0,0 +1 @@
+084	2026-07-08	МТС (Маркетолог)	Senior Frontend developer	Evaluated	1.0/5	❌	[084](reports/084-mts-marketolog-2026-07-08.md)	Remote ок, но React/Next.js (не Vue) + Senior 6+ лет — стек и уровень не совпадают
@@ -0,0 +1 @@
+085	2026-07-08	МТС (GOODOK)	Middle Flutter Developer	Evaluated	2.0/5	❌	[085](reports/085-mts-goodok-2026-07-08.md)	Гибрид/удалённо ок, BLoC/Dart совпадают, но глубина (Isolates, локальные БД, тесты) и стаж 3-6 лет vs ~1 год — разрыв
@@ -0,0 +1 @@
+086	2026-07-08	МТС (MWS Tables)	Fullstack developer React/Node.JS	Evaluated	1.0/5	❌	[086](reports/086-mts-tables-2026-07-08.md)	Гибрид в Москве — локационный блокер (Тула, без релокации); плюс стек/стаж выше профиля
@@ -0,0 +1 @@
+087	2026-07-08	МТС (MWS Octapi)	Стажер Fullstack QA	Evaluated	1.0/5	❌	[087](reports/087-mts-octapi-2026-07-08.md)	Гибрид в Москве — локационный блокер; плюс QA-функция и стажёрский уровень
@@ -0,0 +1 @@
+088	2026-07-08	МТС (CICADA8)	Junior Fullstack-аналитик	Evaluated	1.0/5	❌	[088](reports/088-mts-cicada8-2026-07-08.md)	Офис в Москве — локационный блокер; плюс роль аналитика, не разработчика
@@ -0,0 +1 @@
+089	2026-07-08	МТС (Телеком платформа)	Middle QA fullstack	Evaluated	1.0/5	❌	[089](reports/089-mts-telecom-platforma-2026-07-08.md)	Remote ок, но это QA-роль (тестирование), не разработка
@@ -0,0 +1 @@
+090	2026-07-08	МТС	Middle/Senior Fullstack разработчик	Evaluated	1.0/5	❌	[090](reports/090-mts-fullstack-2026-07-08.md)	Единственная из 7 без явного remote-тега (только Москва+гибкий график) — консервативно офис, блокер
@@ -0,0 +1 @@
+091	2026-07-09	Неизвестный работодатель (TG-брокер)	WebView-разработчик (Unity/Flutter/Native)	Evaluated	1.0/5	❌	[091](reports/091-webview-broker-2026-07-09.md)	Сдельная оплата, нет компании, не соответствует профилю — не рекомендуется
@@ -0,0 +1 @@
+092	2026-07-09	Swiftly / STA TECH	Junior Flutter-разработчик	Evaluated	1.2/5	❌	[092](reports/092-swiftly-sta-tech-2026-07-09.md)	Стек совпадает (Flutter), но без оплаты до релиза и монетизации — не рекомендуется
@@ -0,0 +1 @@
+093	2026-07-09	АО ГК Орбита	Fullstack C# Senior	Evaluated	1.0/5	❌	[093](reports/093-gk-orbita-2026-07-09.md)	C# стек, не пересекается ни с одним архетипом кандидата
@@ -0,0 +1 @@
+094	2026-07-09	EvaTeam	Frontend разработчик Angular	Evaluated	1.3/5	❌	[094](reports/094-evateam-2026-07-09.md)	Angular, не Vue.js — стековое несовпадение
@@ -0,0 +1 @@
+095	2026-07-09	TOOLUM	Fullstack разработчик Next.js	Evaluated	1.3/5	❌	[095](reports/095-toolum-2026-07-09.md)	Next.js/React, не Vue.js — стековое несовпадение
@@ -0,0 +1 @@
+096	2026-07-09	AppDepo	Flutter Developer (Junior)	Evaluated	3.2/5	✅	[096](reports/096-appdepo-2026-07-09.md)	Сильное совпадение по Flutter/Bloc/Firebase, но зарплата не раскрыта (почасовая) — уточнить ставку до отклика
@@ -0,0 +1 @@
+097	2026-07-09	Awelite	Fullstack Web разработчик	Evaluated	1.3/5	❌	[097](reports/097-awelite-2026-07-09.md)	PHP-компания, бэкенд-стек не совпадает с профилем (Node.js)
@@ -0,0 +1 @@
+098	2026-07-09	Young&Junior / СОФИ И ПАРТНЕРЫ	Junior Frontend Разработчик	Evaluated	1.6/5	❌	[098](reports/098-young-junior-2026-07-09.md)	React (не Vue), ЗП от 90к ниже минимума кандидата, несовпадение брендинга работодателя
@@ -0,0 +1 @@
+099	2026-07-09	Constructor TECH	Frontend Engineer (Intern)	Evaluated	1.0/5	❌	[099](reports/099-constructor-tech-2026-07-09.md)	Стажировка только для студентов Constructor University/Nexademy/Nexford — формальный барьер входа, вероятно неоплачиваемая
@@ -0,0 +1 @@
+60	2026-07-07	ЗАО КонсультантПлюс	Веб разработчик Vue/Js	Evaluated	3.0/5	❌	[060](reports/060-konsultantplyus-2026-07-07.md)	Дубликат URL из сегодняшнего batch-инпута (тот же t.me/s/runello_rus_javascript/1448); отчёт 060 от 2026-07-07 уже покрывает эту вакансию — новый отчёт не создан. PDF не был фактически сгенерирован несмотря на балл=порог 3.0 (только HTML), требует ручной генерации при необходимости.
@@ -0,0 +1 @@
+100	2026-07-09	CHOICY (агентство, B2C SaaS партнёр)	Senior Full-stack React Developer	Evaluated	1.0/5	❌	[100](reports/100-choicy-react-b2c-saas-2026-07-09.md)	8+ лет опыта требуется + React, не Vue — двойное несовпадение
@@ -0,0 +1 @@
+101	2026-07-09	Tilda Publishing	Head of Frontend-разработки	Evaluated	1.0/5	❌	[101](reports/101-tilda-publishing-2026-07-09.md)	Управленческая роль (15+ подчинённых, 3+ года лидерства) + гибрид Москва/СПб — двойной блокер
@@ -0,0 +1 @@
+102	2026-07-09	Data World (аутстафф Сбер)	Frontend разработчик (опыт работы в банке)	Evaluated	1.4/5	❌	[102](reports/102-data-world-2026-07-09.md)	React + мин. 3 года опыта — двойное несовпадение, несмотря на хорошую ЗП и удалёнку
@@ -0,0 +1 @@
+103	2026-07-09	CORE TECHNOLOGY	Frontend Developer (Middle)	Discarded	1.0/5	❌	[103](reports/103-core-technology-2026-07-09.md)	Hard blocker: only on-site Tashkent, no remote, relocation required
@@ -0,0 +1 @@
+104	2026-07-09	Lunas.pro	Lead Graphics Engineer - Web 3D	Discarded	1.0/5	❌	[104](reports/104-lunas-pro-2026-07-09.md)	Skill mismatch: WebGPU/Three.js/shaders, no overlap with Vue/Flutter background; also Lead vs Middle level
@@ -0,0 +1 @@
+105	2026-07-09	Podscribe	Senior Full-stack Engineer (Node.js)	Discarded	1.5/5	❌	[105](reports/105-podscribe-2026-07-09.md)	Senior level, React not Vue, Python/ClickHouse gap, B1 English risk on intl team
@@ -0,0 +1 @@
+106	2026-07-09	IDT	Senior Frontend Software Engineer (React)	Discarded	1.0/5	❌	[106](reports/106-idt-2026-07-09.md)	Employment tied to Georgia/Moldova/Belarus, not RU; also React/Electron/AWS CDK stack mismatch
@@ -0,0 +1 @@
+107	2026-07-09	Acronis	Senior Full-Stack Developer	Discarded	1.0/5	❌	[107](reports/107-acronis-2026-07-09.md)	Bulgaria hybrid relocation + Senior 6+yrs gap; strong RBAC/Vue content match though
				`@@ -0,0 +1 @@`
				Resumen: Esta oferta (`t.me/s/runello_rus_javascript/1445`) es un duplicado exacto del post ya evaluado como reporte 059 el 2026-07-07 — mismo empleador anónimo, mismo stack (React + Python/LLM/LangChain/RAG), mismo bot de Runello como único canal de aplicación. Score 1.0/5, legitimidad Suspicious (empleador no revelado + aplicación forzada vía bot de terceros), sin match real con el perfil Vue/Flutter de la candidata. No se generó PDF (por debajo del umbral 3.0). Reporte guardado en `reports/074-unknown-employer-runello-2026-07-08.md` y línea de tracker en `batch/tracker-additions/9.tsv`. Recomendación: no aplicar.
				`@@ -0,0 +1 @@`
				`001 2026-07-06 Яндекс (Лавка) Разработчик фронтенда (Vue/React) в Лавку Evaluated 1.0/5 ❌ [001](reports/001-yandex-lavka-2026-07-06.md) Сильный стек-мэтч (Vue3+тесты+legacy+ритейл), но офис/гибрид в Москве/СПб без удалёнки — жёсткий локационный блокер`
				`@@ -0,0 +1 @@`
				`002 2026-07-06 VK Frontend-разработчик (MAX) Evaluated 2.3/5 ❌ [002](reports/002-vk-max-2026-07-06.md) Удалёнка доступна и уровень Middle совпадает, но ядро роли — React-экосистема, которой нет в CV (стек кандидата — Vue)`
				`@@ -0,0 +1 @@`
				`003 2026-07-06 VK Старший Frontend-разработчик сообщества (ВКонтакте) Evaluated 1.8/5 ❌ [003](reports/003-vk-vkontakte-senior-2026-07-06.md) Двойной разрыв: требуется Senior/4+ лет и React, у кандидата Middle/~2.5 года и стек Vue; не рекомендуется откликаться`
				`@@ -0,0 +1 @@`
				`004 2026-07-06 Ozon Frontend-разработчик (ERP и учётные системы) Evaluated 4.2/5 ✅ [004](reports/004-ozon-2026-07-06.md) Лучший фит пакета: Middle Vue3/TS, удалённо, formal "3+ года JS" порог выше факта (~2 года)`
				`@@ -0,0 +1 @@`
				`005 2026-07-06 Ozon Стажёр Frontend-разработчик (Маркетинговая платформа) Evaluated 2.5/5 ❌ [005](reports/005-ozon-2026-07-06.md) Уровень "стажёр" ниже Middle-профиля кандидата`
				`@@ -0,0 +1 @@`
				`006 2026-07-06 Wildberries Frontend-разработчик (Платформа) Evaluated 2.8/5 ❌ [006](reports/006-wildberries-2026-07-06.md) Стек React, не Vue — adjacent архетип, не primary`
				`@@ -0,0 +1 @@`
				`007 2026-07-06 Wildberries (RWB) Senior Frontend developer (Angular) в WB Job Evaluated 1.7/5 ❌ [007](reports/007-wildberries-2026-07-06.md) Удалённо, но нет Angular и стаж вдвое короче требуемых 4 лет — строго не рекомендуется`
				`@@ -0,0 +1 @@`
				`008 2026-07-06 Т-Банк Frontend-разработчик — React Evaluated 2.0/5 ❌ [008](reports/008-tbank-react-2026-07-06.md) Удалённо/гибко, но нет коммерческого React (обязательно от 4 лет) — строго не рекомендуется`
				`@@ -0,0 +1 @@`
				`009 2026-07-06 Т-Банк Frontend-разработчик — Angular Evaluated 2.3/5 ❌ [009](reports/009-tbank-angular-2026-07-06.md) Удалённо/гибко, Middle/Senior, но нет Angular (обязательно от 2 лет) — строго не рекомендуется`