Privilege log from discovery records

You are preparing structured source data for a draft privilege log. This is document administration, not legal advice. Do not decide whether privilege applies.

Read the reviewed discovery records below. Return one JSON object per record in a JSON array. Use exactly these fields:
- record_id
- bates_start
- bates_end
- document_date
- document_type
- author
- recipients
- copied_parties
- subject_or_title
- family_id
- parent_id
- attachment_ids
- reviewer_privilege_designation
- reviewer_claimed_basis
- reviewer_notes
- source_reference
- missing_fields
- ambiguity_notes

Rules:
1. Copy facts only from the supplied records. Do not infer a person’s role, legal capacity, relationship or purpose.
2. Preserve names, dates and identifiers as given. If a field is absent, use null.
3. Put every absent field that matters for a log entry in missing_fields.
4. Put conflicting values, unclear speaker identities and unclear family relationships in ambiguity_notes.
5. If the same record appears more than once, retain both entries and state the conflict in ambiguity_notes.
6. Do not add a privilege basis where the reviewer has not supplied one.

Reviewed discovery records:
[paste reviewed discovery export, reviewer notes or record summaries here]

Create a draft privilege log from the normalised discovery records below. This is not legal advice. A qualified lawyer must decide any claimed basis and approve every entry.

Return a Markdown table with one row per potentially withheld document or document family. Use exactly these columns:
| log_id | bates_range | document_date | document_type | author | recipients | copied_parties | neutral_description | reviewer_claimed_basis | factual_basis_from_record | family_treatment | decision_required | missing_information | source_reference |

Instructions:
1. Include a record only if reviewer_privilege_designation, reviewer_claimed_basis or reviewer_notes indicates potential privilege, work product, legal advice, litigation purpose, or a request for legal input.
2. In reviewer_claimed_basis, reproduce the reviewer’s stated label. If none exists, write UNRESOLVED. Do not select a legal basis yourself.
3. Write neutral_description from the supplied metadata and notes. Do not reveal the substance of legal advice or privileged communications.
4. In factual_basis_from_record, state only the recorded facts that may support the reviewer’s label, such as author, recipient, stated purpose or reviewer annotation. If there are no such facts, write NOT STATED.
5. Use family_treatment to identify whether the entry is a parent, attachment, duplicate, or incomplete family. Do not assume that a parent’s treatment applies to attachments.
6. Put a clear lawyer question in decision_required whenever the designation, participants, purpose, date, family relationship or scope is unclear.
7. List missing dates, recipients, roles, attachments and source gaps in missing_information.
8. After the table, add a section called Excluded records listing record_id and the factual reason each record was not included.

Normalised records:
[paste JSON array of normalised records here]

Review the document families below for a draft privilege log. This is not legal advice. Do not state that any document is privileged or not privileged.

Return a Markdown table with exactly these columns:
| family_id | record_id | relationship | available_metadata | possible_log_treatment | risk_or_gap | lawyer_decision_needed |

Apply these rules:
1. Treat every attachment as a separate record unless the supplied data explicitly says it is unavailable or non-responsive.
2. Identify parent-child links only where family_id, parent_id, attachment_ids, file names or supplied notes support the link.
3. Mark duplicate or near-duplicate records separately. State the source evidence for the duplicate finding.
4. In possible_log_treatment, use one of: separate entry, consider grouping, duplicate review, incomplete family, or unresolved.
5. Do not carry a reviewer designation from an email to an attachment without an explicit record-level designation or stated factual basis.
6. Where a family is incomplete, identify the missing record or metadata needed.
7. Make lawyer_decision_needed a precise question, not a general warning.

Document family data:
[paste parent, attachment and duplicate records here]

Audit the draft privilege log against the reviewed discovery records below. This is a quality-control exercise, not legal advice. Do not resolve legal questions.

Return two Markdown tables.

First table heading: Conflicts and omissions
Use these columns:
| issue_id | log_id_or_record_id | issue_type | evidence_from_source | effect_on_draft_log | action_for_lawyer_or_review_team |

Second table heading: Records needing a decision
Use these columns:
| record_id | bates_range | current_reviewer_designation | information_available | exact_decision_needed |

Check specifically for:
- log entries with no source record;
- source records marked for potential privilege but absent from the draft log;
- conflicting author, recipient, date, Bates range or document type values;
- an asserted basis with no supporting reviewer note or factual metadata;
- descriptions that disclose substantive advice rather than document characteristics;
- unclear document families, attachments and duplicates;
- missing source references;
- records that combine multiple documents or inconsistent Bates ranges.

For every finding, quote the relevant supplied field value. If evidence is absent, say NOT PRESENT IN SUPPLIED DATA. Do not fill gaps from general knowledge.

Draft privilege log:
[paste draft log here]

Reviewed discovery records:
[paste source records or normalised records here]

Turn the unresolved privilege-log issues below into a lawyer decision queue. This is not legal advice and does not decide privilege. It prepares factual questions for a qualified lawyer.

Return:
1. A Markdown table with exactly these columns:
| priority | issue_id | record_id_or_log_id | decision_question | facts_available | facts_missing | proposed_owner | source_reference |
2. A section called Information requests, with numbered requests that can be sent to the review team or document custodian. Each request must name the record ID and the exact missing fact.
3. A section called Ready for log update, listing only issues that can be corrected from supplied factual data without a legal decision.

Rules:
- Use priority values High, Medium or Low. Base priority on whether the gap prevents a complete, accurate or traceable draft entry. Do not prioritise by guessing legal outcome.
- Write decision_question so it can be answered by a lawyer, for example whether the recorded participants and purpose support the reviewer’s proposed treatment.
- Assign proposed_owner only where the material identifies one. Otherwise write UNASSIGNED.
- Do not recommend withholding, production or redaction.
- Do not invent facts, roles, dates or communication purposes.

Unresolved issues and draft log:
[paste audit output, log rows and relevant source records here]