ooligo
claude-skill

E-discovery custodian questionnaire with Claude

Difficulty
intermediate
Setup time
30-60 min
For
legal-ops-manager
Legal Ops

Stack

A Claude Skill that takes a legal-hold scope document and a custodian’s role, and generates a structured custodian interview questionnaire — organized by data locations, communication channels, device inventory, data-retention practices, and role-specific questions — ready for the legal-ops manager to review and send. The questionnaire is not a canned template: it selects question modules based on the matter type (commercial litigation, regulatory, internal investigation) and the custodian’s function (executive, sales ops, finance, engineering, IT, HR), producing a draft specific enough for a real interview rather than a generic checklist.

This page covers when to run the skill, when not to, what it costs, and the named failure modes you should account for before deploying it on production holds.

When to use

Reach for the skill at the start of the data-collection phase — after the legal-hold scope has been documented by counsel and before custodian interviews begin.

  • Litigation holds with multiple custodians. When the custodian list spans five or more people across departments, the time cost of drafting a role-calibrated questionnaire for each one is material. The skill drafts per custodian in under a minute; the legal-ops manager reviews and calibrates in 10-15 minutes per questionnaire rather than 45-60.
  • Regulatory investigations with short timelines. Government requests typically have 30-day response windows. The skill compresses the questionnaire-drafting cycle so the legal-ops team can interview more custodians before the preservation deadline.
  • Internal investigations where consistency matters. When the same set of questions needs to reach all relevant custodians in the same format — important for demonstrating good-faith preservation efforts if the matter later becomes litigation — the skill enforces structural consistency the legal-ops team would otherwise produce manually.
  • Onboarding junior legal-ops staff. New staff often draft questionnaires that miss informal channels (personal email, messaging apps) or role-specific data sources. The skill includes these by construction.

The artifact bundle lives at apps/web/public/artifacts/ediscovery-custodian-questionnaire-skill/ and ships:

  • SKILL.md — the Skill definition with method, output format, and watch-outs
  • references/1-question-library.md — the question library, organized by section and module, with rationale per question
  • references/2-hold-scope-template.md — fillable input template for the hold scope document
  • references/3-custodian-interview-checklist.md — the post-interview checklist the legal-ops manager completes after the custodian call

When NOT to use

The skill is narrow on purpose. Stop before invoking it in any of these cases.

  • Deciding who is a custodian. Whether a specific employee is a legal custodian subject to a preservation obligation is a legal judgment call. The skill drafts interview questionnaires for confirmed custodians; it does not identify the custodian list. That determination is counsel’s. Consult counsel before finalizing your custodian scope.
  • Generating strategy memos. Custodian significance to a defense theory, anticipated privilege claims on custodian communications, and deposition strategy are work product. This skill produces a data-collection questionnaire, not a privileged strategy document.
  • Active privilege review. Documents already collected and queued for privilege review belong in a review platform (Relativity with attorney reviewers, not this skill). The skill is for the intake phase — identifying what exists — not the review phase.
  • Auto-sending questionnaires to custodians. The legal-ops manager reviews and edits every output before sending. The skill does not distribute.
  • HR investigations where dual-purpose privilege rules apply. Employment matters that may become litigation carry distinct privilege considerations for the investigation materials themselves. Consult counsel before using this skill in those contexts.

Setup

  1. Download the bundle from apps/web/public/artifacts/ediscovery-custodian-questionnaire-skill/ and place it in ~/.claude/skills/ (Claude Code) or upload to a Claude.ai project.
  2. Open references/1-question-library.md and update the platform-specific question text for your organization. The default questions reference generic platforms; your custodians work in specific tools (name your CRM, your email archive, your collaboration platform). Platform-specific questions get more accurate responses.
  3. Fill in references/2-hold-scope-template.md once per matter before running the skill on any custodian. The skill requires a structured scope document; an incomplete scope produces questions that may not hold up if the questionnaire is challenged in litigation. Consult counsel on scope finalization.
  4. Run the skill on one test custodian with a closed matter (a completed hold you can verify against). Check that the role modules are correct and that no questions reference platforms the custodian doesn’t use.
  5. For each new custodian: fill in the scope template (or reference the matter’s existing scope document), provide the custodian’s name, title, and role category, and invoke the skill. Review the output before sending.

What the skill actually does

Four steps, in order — and the engineering choices behind them.

Step 1 — Parse the hold scope and custodian context. The skill extracts the matter name, matter type, key issues summary, relevant date range, custodian name, and custodian title from the scope document. If any required field is missing, the skill emits a structured error and stops — it does not generate a questionnaire against an incomplete scope. Why: incomplete scope inputs produce vague questions that don’t prompt the specific answers needed for the hold record, and vague questionnaires are harder to defend if the hold’s good-faith basis is ever challenged.

Step 2 — Select question modules. The skill maps the matter type and custodian role to a specific set of question modules. Base modules run on every questionnaire: data locations, communication channels, device inventory, data-retention practices, and a relevant-date-range check. Role modules add on top (executive, sales ops, finance, engineering, IT, HR). Matter-type modules add further (commercial litigation asks about the specific counterparty; regulatory asks about prior government contact; internal investigation adds escalation-log questions). Why separate modules rather than one giant question list: a questionnaire scoped to the custodian’s role and the matter type is shorter and more focused, which means higher-quality responses. Custodians who receive a 60-question generic questionnaire give shorter, less detailed answers than those who receive a 15-question focused one.

Step 3 — Generate the draft questionnaire. The output is two sections: the custodian-facing questionnaire (numbered sections, response lines beneath each question, a confidentiality header) and a legal-ops notes section (flagged items for counsel only, not shown to the custodian). The confidentiality header is included by default on every output because legal-ops managers drafting ad hoc frequently omit it — the skill includes it by construction; counsel deletes it only with intention. Consult counsel on privilege framing for the specific matter.

Step 4 — Flag high-risk items. After the questionnaire, the skill appends a legal-ops notes section marking questions that touch privileged materials, questions referencing named third parties, and any data-retention answers that may indicate spoliation risk. This section is for the legal-ops manager and counsel only.

Cost reality

Per-questionnaire token cost is low because the inputs are compact (hold scope + custodian role = typically 500-2,000 input tokens; question library = ~4,000 input tokens; output = 1,500-3,000 tokens depending on module count). At current Claude pricing ($3/M input tokens, ~$15/M output tokens on the cost-effective tier):

  • Typical custodian questionnaire (3 base modules + 1 role module, 20 questions): ~$0.03-0.05 per questionnaire.
  • Executive with full module set (base + executive + matter-type overlays, 30+ questions): ~$0.06-0.10 per questionnaire.
  • 50-custodian regulatory investigation: ~$2-5 in token spend.

The non-token cost dominates: 10-15 minutes of legal-ops manager review time per questionnaire is the real floor. That’s the irreducible minimum — the skill is not a replacement for review, it is a replacement for the 45-60 minutes of drafting. The net saving on a 20-custodian hold is roughly 15-20 attorney-equivalent hours.

Relativity or your e-discovery platform is the downstream consumer for what the questionnaires surface — custodian data sources feed the collection scope, which feeds the processing queue in Relativity. Budget Relativity processing costs separately; token spend on questionnaire generation is not the significant cost item.

Success metric

Two metrics worth tracking from the first hold.

  • Coverage completeness. After the custodian interviews, compare the data sources custodians disclosed to the sources your IT team independently identifies in the data map. The gap is your under-reporting rate. A well-calibrated questionnaire with explicit informal-channel questions should produce a disclosure rate within 10-15% of the IT data map. Gaps above 25% mean the question library needs stronger platform-specific language for your organization.
  • Interview duration. A focused, role-specific questionnaire should take 20-30 minutes per custodian over video call. Interviews running 50+ minutes typically mean the questionnaire is asking redundant questions or covering topics the custodian can’t answer without IT support — trim accordingly.

vs alternatives

vs manual questionnaire drafting by the legal-ops manager. Manual drafting from a prior-matter template takes 45-60 minutes per custodian and produces questionnaires calibrated to the prior matter, not the current one. The most common failures: informal channels (Signal, WhatsApp, personal email) are omitted from the communications section; role-specific questions (pricing approvals for sales ops, version-control repositories for engineering) are missing. This skill drafts per-matter, per-role in under a minute and includes these by construction.

vs Relativity built-in custodian modules. Relativity’s Data Ingestion and Custodian Portal features handle chain-of-custody tracking and collection authorization — they don’t draft interview questionnaires. The skill and Relativity are complementary: the skill generates the questionnaire, the interview produces the data source list, Relativity executes the collection. If your organization is fully in Relativity and uses its custodian-portal workflow, use this skill for the interview phase and Relativity for the collection phase.

vs Logikcull, Everlaw, or Exterro end-to-end e-discovery platforms. These platforms include custodian-questionnaire modules as part of a broader e-discovery workflow. If you’re already licensed on one of them and they cover your matter type, use the built-in module — setup cost is lower. This skill is the right choice when you don’t have an e-discovery platform with questionnaire functionality, when you need more question-module flexibility than the platform’s template allows, or when you’re handling a matter type the platform doesn’t cover well (internal investigations, especially).

Watch-outs

  • Scope creep into legal advice. The questionnaire collects factual information about data locations; it does not ask custodians to characterize their preservation obligations. Guard: the skill explicitly excludes questions of the form “Did you know you were required to preserve X?” — these are legally loaded and are stripped from the question library at authoring time. The output never contains them.
  • Omission of informal channels. Custodians underreport personal email, SMS, and encrypted messaging apps unless asked by name. Guard: the base communication-channels module lists platforms by name (WhatsApp, Signal, iMessage, Telegram, personal Gmail) rather than generically. If data_sources_known includes no informal channels, the skill adds a [completeness-flag] in the legal-ops notes section alerting the manager to probe further.
  • Privilege header omission. Ad hoc questionnaires often lack the confidentiality header, which can complicate work-product assertions if the questionnaire is later produced. Guard: the skill includes the header by default on every output; counsel can remove it with intentional deletion, not by accident.
  • Spoliation risk not surfaced. A custodian who reveals a 30-day auto-delete setting on their primary email account is a spoliation risk if the hold predates the interview. Guard: the data-retention module is on every questionnaire, and any retention answer mentioning deletion or auto-expiry during the relevant period is flagged [spoliation-watch] in the legal-ops notes section, prompting immediate escalation to counsel.
  • Stale question library. The references/1-question-library.md template covers generic platform categories. If your organization switches CRMs, collaboration tools, or cloud storage and the library isn’t updated, the questions reference platforms custodians don’t use. Guard: the skill checks the last_updated field in the question library and emits a warning when it is more than 12 months old, prompting the legal-ops manager to review for platform accuracy.

Stack

  • Claude — question-module selection, questionnaire generation, legal-ops flag generation
  • Relativity — downstream e-discovery platform for the collection and review phase that the custodian interviews feed into

Related: the apps/web/public/artifacts/ediscovery-custodian-questionnaire-skill/ bundle ships three reference files. The question library (references/1-question-library.md) is the primary adapter — customizing it for your organization’s platforms is the highest-ROI configuration step.

Files in this artifact

Download all (.zip)