Editorial methodology
How CasePrompts content is built and maintained
CasePrompts publishes reusable prompt text, not legal advice. This page explains the technical and editorial controls behind the library so readers and automated systems can evaluate what a published prompt represents.
Record-first control loop
From source to reviewed work product
CITE defines the assignment and evidence standard, but the source and the human reviewer remain the authority at either end of the workflow.
- 01 / RECORD47:12–48:3
Start with the source
Identify the deposition, exhibit, medical record, or production that controls the assignment.
- 02 / CITE ASSIGNMENTCITE
Define the work
Set the context, instructions, output template, and evidence rule before asking for analysis.
- 03 / LOCATED OUTPUT
Keep the cite beside it
Require each material finding to carry a page, line, Bates, paragraph, or record-page location.
- 04 / HUMAN VERIFICATION
Review against the record
A person checks the source, revises the draft, and decides what becomes work product.
Deterministic CITE assembly
Library prompts use four named blocks: Context, Instructions, Template, and Evidence. The builder selects a published template, inserts the values entered in the browser, resolves the citation instruction for the selected source type and destination, and optionally composes a published practice lens. The assembly function is deterministic: the same template version and inputs produce the same prompt text. It does not call a language model.
Published and draft governance
Every template and practice lens has an explicit published or draft status. Public library, practice, prompt, sitemap, and machine-readable discovery pages query published entries only. Draft entries are excluded from those surfaces until their status is deliberately changed.
Published means
The entry passed the repository’s current content checks, has a published status, and is deliberately included in public routes and discovery files.
Published does not mean
Lawyer-reviewed, approved for a jurisdiction, substantively correct, or sufficient for a firm’s confidentiality, security, ethics, or professional requirements.
Users must verify source citations and substantive output against the record and follow their own professional and organizational requirements.
The prompting guides and workflow hubs received an internal editorial review dated on each page. They have not received independent legal or ethics review. The privacy policy and terms describe current intended behavior but remain plain-language legal drafts and are marked noindex until independent review is complete.
Schema and content validation
Templates and lenses are stored as structured content and parsed against versioned Zod schemas. Project validation checks relationships that a single file cannot prove on its own—for example, that a workflow references a published template, uses a compatible document type, and does not select an excluded lens. Invalid content fails validation instead of silently appearing on the site.
Versioning and tests
Each template and lens carries an integer version. Template versions are intended to be incremented when their CITE blocks change; lens versions cover changes to their emphases and workflows. Automated tests cover interpolation, required and optional variables, citation rules, exact CITE labels, lens composition and exclusions, workflow relationships, schema constraints, and the rule that drafts do not reach public routes.
Corrections and provenance
The content files in the CasePrompts repository are the source of truth for prompt text, status, and version. Corrections are made there so the builder, public pages, and generated discovery files stay aligned. To report an error, include the page URL and a short description and email support@caseprompts.com.
Record-type workflow pages disclose their editorial review date, scope, source-coordinate conventions, and human verification checks. Those pages offer general orientation and intentionally avoid jurisdiction-specific sufficiency or admissibility conclusions.
AI-tool information
Pages about third-party AI tools explain how to use CasePrompts output with those tools. Product names, features, policies, and privacy terms can change. Treat those pages as practical orientation. Each page now shows its product-source review date and links the first-party documentation used for capability and data-handling statements. A dated source review is not a security assessment or an ongoing guarantee; confirm the exact product, plan, settings, contract, and provider documentation before use.