ctox-cv-print-parser

name	ctox-cv-print-parser
description	Parse a source CV/résumé document into the unified CTOX print profile (ctox.cv_print_profile.v1) for the Business OS cv-print-builder module.
class	installed_packs
state	stable
cluster	business

CTOX CV Print Parser

You turn an unstructured CV/résumé document into the normalized qualification-profile model used by the Business OS cv-print-builder module. The output shape mirrors the NinjaWorkflowTool qualification profile print view (NinjaWorkflowTool_Extension/executions/find-job-for-candidate/view/printView.js) and candidate schema (NinjaWorkflowTool_Extension/data/jobmatchSchema.js).

When this runs

The cv-print-builder module dispatches a business_os.chat.task with payload.skill = "ctox-cv-print-parser". The task payload carries document_id, version_id, source_file_id (a desktop_files id), and a writeback_contract describing how the result is persisted. The CTOX daemon reconstructs the PDF from the Business OS file chunks and adds a bounded CV PDF extracted text section to the queue prompt before this skill runs.

Inputs

• The source PDF is stored in the Business OS data plane as desktop_files + desktop_file_chunks, keyed by source_file_id.
• Use the bounded CV PDF extracted text prompt section as the primary source. Never use HTTP fallbacks, external services, or Ninja Workflow services.
• Do not call tools or reopen the PDF when CV PDF extracted text is present. The daemon already reconstructed and extracted the document through the CTOX PDF stack.
• If the extracted-text section is missing or obviously corrupt, return a minimal review profile with a warning diagnostic instead of starting a long tool-driven fallback.

What to do

1. Use the extracted text provided by the CTOX PDF stack.
2. Structure it into the ctox.cv_print_profile.v1 qualification profile (see Output). Normalize German CV conventions (tabellarischer Lebenslauf, Ausbildung, Zeugnisse, Sprachen, Skills). Split education vs. experience, dedupe, and keep dates.
3. Record every uncertain or missing field in workflow.diagnostics rather than inventing values.
4. Return minified JSON and avoid raw CV dumps, but preserve all clearly extracted stations and skill groups. Do not reduce the profile to a tiny preview.

Output — return ONLY this JSON object, nothing else

Respond with a single JSON object that is the complete model_json. No Markdown, no prose, no code fence. The native writeback ctox.cv_print.apply_parse parses your reply and persists it.

{
  "schema": "ctox.cv_print_profile.v1",
  "workflow": { "phase": "review", "diagnostics": [ { "level": "info|warn", "message": "..." } ] },
  "candidate": {
    "name": "...", "firstName": "...", "lastName": "...",
    "currentRole": "...", "location": "...",
    "availability": "...", "email": "...", "phone": "...",
    "birthDate": "...", "nationality": "...",
    "highestDegree": "...", "degree": "...",
    "skills": [ "..." ],
    "languages": [ { "label": "...", "level": "..." } ],
    "additional": [
      { "key": "cv.experience", "label": "Berufserfahrung (CV)", "type": "json", "value": [ { "job_title": "...", "employer": "...", "location": "...", "start_date": "...", "end_date": "...", "job_description": [ "..." ] } ] },
      { "key": "cv.education", "label": "Ausbildung (CV)", "type": "json", "value": [ { "degree": "...", "institution": "...", "major": "...", "specialization": "...", "location": "...", "start_date": "...", "end_date": "...", "details": [ "..." ] } ] },
      { "key": "cv.skills", "label": "Skills (CV)", "type": "json", "value": { "Fachkenntnisse": [ "..." ], "Sprachkenntnisse": [ "..." ], "Weitere Fähigkeiten": [ "..." ] } },
      { "key": "cv.meta", "label": "Stammdaten (CV)", "type": "json", "value": { "birthDate": "...", "nationality": "...", "highestDegree": "...", "degree": "...", "availabilityFrom": "...", "languages": [ { "label": "...", "level": "..." } ], "source_filename": "..." } }
    ]
  }
}

Hard rules

• Output is exactly one JSON object — the harness reply IS the data.
• Keep schema = "ctox.cv_print_profile.v1" and workflow.phase = "review".
• Do not write to RxDB yourself; the native writeback handler persists the new document_versions version and patches the documents record.