← Files LuciaARCHIVED FILE
privacy/workstreams/revisione-documentale.json
7.8 KB · Nov 21, 2026 · 00:11 UTC
{"display_name":"Revisione documentale","external_boundaries":[],"governed_paths":["skills/revisione-documentale","scripts/legal_citations.py","scripts/legal_citations.schema.json","scripts/legal_citations_labels.json","scripts/legal_documents.py","scripts/legal_documents_labels.json","scripts/legal_documents_review.schema.json","scripts/legal_docx.py","scripts/legal_matter.py","scripts/legal_matter.schema.json","scripts/legal_matter_labels.json","scripts/legal_negotiation.py","scripts/legal_negotiation.schema.json","scripts/legal_negotiation_labels.json","scripts/legal_playbook.py","scripts/legal_playbook.schema.json","scripts/legal_playbook_labels.json","scripts/legal_proofreading.py","scripts/legal_proofreading.schema.json","scripts/legal_proofreading_labels.json","scripts/legal_word_bridge.py","skills/revisione-contratti/references","skills/lucia/references/model-data-report-contract.md"],"model_context":{"classes":[{"content":"Selected text is interpreted by the conversation’s model; document helpers run locally.\n\nThe model may read selected document text, filenames, relationships between agreements and schedules, review questions, quotations and table cells. Chunking limits material read at one time; it does not anonymize data or guarantee that only a summary reaches the model. When selected for this review, passages from firm instructions and legal sources read by the model also enter its context; the report retains references and unresolved checks.\n\nBundled helpers extract text and store copies, references and outputs in the selected folder. They call no document service, Mike server or separate model API and perform no online research. Content read in the conversation is processed by the host model under the account in use; it is not automatically anonymized. Scanned PDFs and unextracted elements remain explicit gaps.","id":"selected-task-evidence","purpose":"Carry out revisione-documentale using the selected evidence and user instructions","runtime_profiles":["openai-codex","anthropic-cowork","openai-chatgpt","google-antigravity"]}],"policy":"real_case_data_may_enter_selected_runtime_model_context"},"packaged_for_runtime":"openai-codex","public_explanations":{"revisione-documentale":{"copy":{"de":["Ausgewählter Text wird vom Modell des Gesprächs interpretiert; Dokumentenhelfer laufen lokal.","Das Modell kann ausgewählte Texte, Dateinamen, Beziehungen zwischen Verträgen und Anlagen, Prüffragen, Zitate und Tabellenzellen lesen. Abschnittsweises Lesen begrenzt die jeweils gelesene Menge; es anonymisiert keine Daten und beschränkt den Modellkontext nicht auf Zusammenfassungen. Für diese Prüfung ausgewählte und gelesene Passagen aus Kanzleivorgaben und Rechtsquellen gelangen ebenfalls in den Modellkontext; der Bericht enthält Quellenangaben und offene Prüfungen.\n\nDie enthaltenen Helfer extrahieren Text und speichern Kopien, Verweise und Ergebnisse im gewählten Ordner. Sie rufen weder Dokumentendienste noch Mike-Server oder separate Modell-APIs auf und recherchieren nicht online. Im Gespräch gelesene Inhalte werden vom Host-Modell unter dem verwendeten Konto verarbeitet, ohne automatische Anonymisierung. Gescannte PDFs und nicht extrahierte Elemente bleiben ausdrücklich als Lücken sichtbar."],"en":["Selected text is interpreted by the conversation’s model; document helpers run locally.","The model may read selected document text, filenames, relationships between agreements and schedules, review questions, quotations and table cells. Chunking limits material read at one time; it does not anonymize data or guarantee that only a summary reaches the model. When selected for this review, passages from firm instructions and legal sources read by the model also enter its context; the report retains references and unresolved checks.\n\nBundled helpers extract text and store copies, references and outputs in the selected folder. They call no document service, Mike server or separate model API and perform no online research. Content read in the conversation is processed by the host model under the account in use; it is not automatically anonymized. Scanned PDFs and unextracted elements remain explicit gaps."],"es":["El modelo de la conversación interpreta el texto elegido; las herramientas documentales funcionan localmente.","El modelo puede leer textos seleccionados, nombres de archivos, relaciones entre acuerdos y anexos, preguntas, citas y celdas. Leer por bloques limita el contenido examinado cada vez; no anonimiza los datos ni garantiza que el modelo reciba solo un resumen. Los pasajes seleccionados de las instrucciones del despacho y fuentes jurídicas que lea el modelo también entran en su contexto; el informe conserva referencias y comprobaciones pendientes.\n\nLas herramientas incluidas extraen texto y guardan copias, referencias y resultados en la carpeta elegida. No llaman a servicios documentales, servidores Mike ni API de modelo separadas, ni investigan en línea. El modelo anfitrión procesa el contenido leído en la conversación según la cuenta utilizada, sin anonimización automática. Los PDF escaneados y elementos no extraídos permanecen como lagunas explícitas."],"fr":["Le texte choisi est interprété par le modèle de la conversation ; les outils documentaires fonctionnent localement.","Le modèle peut lire les textes sélectionnés, noms des fichiers, liens entre accords et annexes, questions, citations et cellules. La lecture par blocs limite le contenu lu à la fois ; elle n’anonymise pas les données et ne garantit pas que seul un résumé parvienne au modèle. Lorsqu’ils sont sélectionnés pour cette revue, les passages lus des instructions du cabinet et des sources juridiques entrent aussi dans le contexte du modèle ; le rapport conserve références et vérifications ouvertes.\n\nLes outils inclus extraient le texte et conservent copies, références et résultats dans le dossier choisi. Ils n’appellent ni service documentaire, ni serveur Mike, ni API de modèle distincte et n’effectuent pas de recherche en ligne. Les contenus lus dans la conversation sont traités par le modèle hôte selon le compte utilisé, sans anonymisation automatique. Les PDF numérisés et éléments non extraits restent des lacunes explicites."],"it":["Il testo selezionato viene interpretato dal modello della conversazione; gli helper documentali operano localmente.","Il modello può leggere il testo dei documenti selezionati, nomi dei file, relazioni tra accordi e allegati, domande della revisione, citazioni e celle della tabella. Il lavoro a blocchi limita il materiale letto per volta, ma non anonimizza i dati o garantisce che solo il riepilogo arrivi al modello. Se selezionati per questa revisione, anche i passaggi delle istruzioni dello studio e delle fonti giuridiche letti dal modello entrano nel contesto; il report conserva riferimenti e verifiche ancora aperte.\n\nGli helper inclusi estraggono testo e conservano copie, riferimenti e risultati nella cartella scelta. Non chiamano servizi documentali, server Mike o API modello separate e non eseguono ricerche online. I contenuti letti nella conversazione sono trattati dal modello ospite secondo l’account utilizzato; non sono anonimizzati automaticamente. I PDF scansionati e gli elementi non estratti restano lacune visibili."]},"status":"relevant"}},"review":{"basis":"external_boundary_review_of_workflow_source","reviewed_at":"2026-11-04","reviewed_by":"privacy-surface-review","source_fingerprint":"9f21eadac6d120a04f351dc63d20240bd4372f533eff08440d2df582fed7ad1e"},"role":"workflow","runtime_profiles":["openai-codex","anthropic-cowork","openai-chatgpt","google-antigravity"],"schema_version":3,"security_controls":[],"service_ids":["managed-python-runtime","operational-incidents","run-receipt-stamping","workflow-usage"],"source_contract_sha256":"6888e2d1cfb1e5c9de5a88c2cbf8e3d3b35ecb3c43432cc237216b01ef815007","workstream":"revisione-documentale"}
SHA-256: ac18388ec27ea06e781c5cf26692b3cef46dd8e92ab7f5b5c5001b77886541cb