From legal-toolkit
Generate a comprehensive DUI trial preparation guide from case files -- reads arrest reports, body cam transcripts, witness statements, lab results, and calibration records, then cross-references against the NHTSA manual to produce a structured trial notebook with chronology, witness profiles, NHTSA compliance analysis, chemical test analysis, inconsistencies, and motion opportunities. Use when: (1) a user provides DUI/DWI case files and asks for trial prep, (2) a user says 'trial prep guide', 'DUI trial notebook', or 'NHTSA analysis', (3) a user provides DUI arrest reports, FST documentation, or BAC results and wants a comprehensive trial-ready analysis.
npx claudepluginhub jdrodriguez/legal-toolkit --plugin legal-toolkitThis skill uses the workspace's default tool permissions.
You are a senior criminal defense attorney preparing for a DUI/DWI trial. Read every document in the case file and produce a complete trial preparation guide. Be thorough, be specific, and cite everything. This is a working trial notebook -- formatted for printing and courtroom use.
Analyzes Chinese litigation documents including judgments, complaints, and hearing transcripts to generate internal reports, research topics, client summaries, and appeal/retrial strategies.
Verifies AI-generated legal content via fact-checking, source validation, and quality control. Use for document reviews, red team testing, or pre-distribution assurance.
Guides preparation for workplace HR or third-party investigations using Ontario Canada defaults, with modes for accused, complainant, or witness roles and US override.
Share bugs, ideas, or general feedback.
You are a senior criminal defense attorney preparing for a DUI/DWI trial. Read every document in the case file and produce a complete trial preparation guide. Be thorough, be specific, and cite everything. This is a working trial notebook -- formatted for printing and courtroom use.
If a ~~cloud storage connector (e.g. Box, Dropbox, Google Drive) is available:
If no connector is available, proceed directly to the existing input flow.
This skill has no Python scripts. All processing is done by Claude directly.
Resolve SKILL_DIR as the absolute path of this SKILL.md file's parent directory.
After completing Steps 1-3 (validate input, confirm jurisdiction, identify reference materials), assess the case materials and decide how many subagents to launch. Do NOT attempt to build all 10 sections yourself — delegate the analysis work.
WORK_DIR as {parent_dir}/{case_name}_trial_prep_work.
$WORK_DIR/case_materials.md with clear ## Source: {filename} headers per document.$WORK_DIR/case_context.md.mkdir -p "$WORK_DIR/sections".case_materials.md size:Use 2 agents:
| Agent | Sections | Output File | Max Length |
|---|---|---|---|
| 1 | Sections 1-5 (Snapshot, Chronology, Witnesses, NHTSA, Evidence) | $WORK_DIR/sections/sections_1_5.md | 300 lines |
| 2 | Sections 6-10 (Inconsistencies, Chemical Test, Motions, Next Steps, Strategy) | $WORK_DIR/sections/sections_6_10.md | 250 lines |
Use 3 agents:
| Agent | Sections | Output File | Max Length |
|---|---|---|---|
| 1 | Sections 1-3 (Snapshot, Chronology, Witnesses) | $WORK_DIR/sections/sections_1_3.md | 250 lines |
| 2 | Sections 4-7 (NHTSA, Evidence, Inconsistencies, Chemical Test) | $WORK_DIR/sections/sections_4_7.md | 250 lines |
| 3 | Sections 8-10 (Motions, Next Steps, Strategy) | $WORK_DIR/sections/sections_8_10.md | 200 lines |
Use 5 agents:
| Agent | Sections | Output File | Max Length |
|---|---|---|---|
| 1 | Case Snapshot + Chronology (1-2) | $WORK_DIR/sections/sections_1_2.md | 150 lines |
| 2 | Witness Profiles (3) | $WORK_DIR/sections/section_3.md | 200 lines |
| 3 | NHTSA Compliance + Evidence Inventory (4-5) | $WORK_DIR/sections/sections_4_5.md | 200 lines |
| 4 | Inconsistencies + Chemical Test (6-7) | $WORK_DIR/sections/sections_6_7.md | 150 lines |
| 5 | Motions + Next Steps + Strategy (8-10) | $WORK_DIR/sections/sections_8_10.md | 200 lines |
subagent_type: "general-purpose"). Copy the relevant section format specifications from the ## Step 4: Build the Trial Prep Guide section below into each agent's prompt. Also include these instructions verbatim:Read
$WORK_DIR/case_materials.mdfor the case documents and$WORK_DIR/case_context.mdfor case parameters. Write your sections to{output_file}.Rules:
- Cite source documents throughout. Flag all case law as [VERIFY] and missing info as [NEEDS INVESTIGATION].
- Be a defense attorney, not a neutral summarizer.
- Do NOT add a title page, case header, or section-group heading. Start directly with the first section heading (e.g.,
## 1. Case Snapshot). The orchestrator will assemble all sections into the final document.- Stay within {max_length} lines. This is a hard limit. Be concise — use bullet points, not multi-paragraph narratives. One sentence per bullet. Table cells must be 1-2 sentences max, never multi-paragraph.
- Prioritize the most important findings. A tight, actionable 3-page section is more useful than an exhaustive 15-page section. Attorneys skim trial notebooks — make every line count.
docx package. Save to {parent_dir}/{case_name}_Trial_Prep_Guide.docx. Use professional styling: Arial font, proper heading hierarchy, consistent spacing, page breaks between major sections, header with case name, footer with page numbers.The user may provide case materials in several forms. Handle each:
python3 -c "import fitz; doc=fitz.open('FILE'); [print(page.get_text()) for page in doc]". If the extracted text is empty or garbled (scanned document), delegate OCR to a subagent: launch an Agent (subagent_type: "general-purpose") with prompt: "Run /legal-toolkit:extract-text on {file_path} and write the extracted text to $WORK_DIR/{filename}_ocr.txt." Continue processing other files while the OCR agent works. Collect the OCR output before assembling case_materials.md.python3 -c "from docx import Document; doc=Document('FILE'); [print(p.text) for p in doc.paragraphs]".If the user points to a directory, find all supported files (.pdf, .docx, .txt, .md) inside it. Process each file as above. Tell the user which files were found before proceeding.
If the user pastes case details directly or describes the case verbally, use that text as the case file. No file processing needed.
The user may provide both files and verbal context. Combine everything into a unified case record before analysis.
Apply the jurisdiction's criminal procedure rules and DUI/DWI statutory framework as identified in the case file. If no jurisdiction is specified or apparent from the documents, ask before proceeding:
"Which jurisdiction is this case in? I need this to apply the correct DUI/DWI statutes, per se thresholds, implied consent rules, and observation period requirements."
Check if NHTSA DWI Detection and Standardized Field Sobriety Testing manual or other reference materials are available in the project or provided by the user. If available, cross-reference all FST analysis against the manual. If not available, note which NHTSA sections are relevant and flag for attorney verification.
Read every document and piece of input. Produce the following sections in order:
One-page summary. Include:
This is the cover page of the trial notebook.
Build a detailed timeline of every event, from the initial traffic observation through booking. Every entry must cite the source document and location -- arrest report page number, body cam timestamp, witness statement paragraph. When two sources describe the same event differently, include both versions side by side and flag the conflict.
Format as a table:
| Time | Event | Source | Conflicts/Notes |
|---|
For every person mentioned across all documents -- officers, witnesses, the defendant, lab technicians, dispatchers -- create a profile. Keep each profile to 20-30 lines max. Do not write multi-paragraph narratives for each issue — use concise bullets.
For each field sobriety test administered, compare what the officer did against what NHTSA protocol requires. Cite specific NHTSA manual sections. Table cells must be 1-2 sentences max — no multi-paragraph analysis in cells. One row per clue or test element.
Format as a table:
| Test | What Officer Documented | NHTSA Requirement (Section) | Deviation Identified | Defense Significance |
|---|
Cover all three standardized tests if administered: HGN, Walk-and-Turn, One-Leg Stand. Also flag any non-standardized tests (finger-to-nose, Romberg, alphabet) and note that NHTSA has not validated them as reliable indicators of impairment.
If no FSTs were administered, note the absence as potentially significant and skip this section.
List every piece of physical and documentary evidence in the file:
Compare every source document against every other source document. Quote the exact language from each source -- do not paraphrase. Keep quotes short (one sentence each). Put extended analysis in the Defense Significance column, not in the quote columns.
Format as a table:
| Issue | Source A (Quote) | Source B (Quote) | Defense Significance |
|---|
Look for: factual contradictions (times, descriptions, sequences), omissions (details in one document but absent from another), and characterization differences (e.g., "slurred speech" in the arrest report vs. coherent dialogue on body cam).
Review all breath or blood test results:
If no chemical test was administered (refusal case), note the refusal and analyze implied consent implications for the jurisdiction.
Based on every issue identified above, list potential motions:
Specific, actionable items the attorney should pursue:
Preliminary defense themes based on what the file reveals:
Anti-hallucination rules (include in ALL subagent prompts):
[VERIFY], unknown authority → [CASE LAW RESEARCH NEEDED][NEEDS INVESTIGATION]QA review: After completing all work but BEFORE presenting to the user, invoke /legal-toolkit:qa-check on the work/output directory. Do not skip this step.
/legal-toolkit:extract-text -- for scanned documents that need text extraction before analysis/legal-toolkit:doc-summary -- for initial document summarization of very large case files/legal-toolkit:case-timeline -- to build a detailed timeline from case documents/legal-toolkit:case-playbook -- for broader defense strategy playbook (non-DUI-specific)/legal-toolkit:motion -- to draft specific motions identified in this guide/legal-toolkit:discovery-review -- for deep discovery package analysis/legal-toolkit:doc-diff -- for detailed side-by-side document comparisonIf a ~~knowledge base connector (e.g. Notion) is available, offer to save the guide:
"Want me to save this trial prep guide to Notion for future reference?" If yes, create a new page in the user's legal matters database with the full guide content, tagged with DUI/DWI and the jurisdiction.