Cross-references a judge's decision, court opinion, or final order against one or more hearing/trial/deposition transcripts to verify whether the record supports each finding. Produces strict dual-citation blocks with document names, page/line references, and verbatim quotes from both sources. Use when preparing appellate review, post-trial motions, record discrepancy audits, or transcript-based fact checks of judicial findings.
Installs into .claude/skills of the current project.
Are you the author of Decision Record Verification?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/thomasmoreai-decision-record-verification)
---
name: decision-record-verification
title: Decision Record Verification
description: Cross-references a judge's decision, court opinion, or final order against one or more hearing/trial/deposition transcripts to verify whether the record supports each finding. Produces strict dual-citation blocks with document names, page/line references, and verbatim quotes from both sources. Use when preparing appellate review, post-trial motions, record discrepancy audits, or transcript-based fact checks of judicial findings.
author: CaseMark
author_url: https://github.com/CaseMark/skills/tree/main/skills/legal/decision-record-verification
license: Apache-2.0
version: 0.1.0
execution_mode: open
jurisdiction: general
practice: litigation
language: en
tags: [analysis, summary]
---
# Decision Record Verification
Cross-reference a court decision against transcripts and surface record alignments, discrepancies, and gaps with exact dual-source citations.
## Prerequisites
- **Decision document** — opinion/order/judgment to test against the record
- **Transcript set** — one or more sessions (trial, hearing, or deposition) with page/line numbering when available
- **Session labels** — clear names for each transcript file (for example, `Day 1 AM`, `Day 1 PM`)
- **Scope confirmation** — identify whether review is full-document or limited to specific findings/issues
If required documents are missing or unlabeled, request them before analysis.
## Quick Start
1. Read the decision in full and extract each factual finding, credibility determination, and testimony-dependent legal conclusion.
2. Convert each finding into a verification claim.
3. Review transcripts in session order and capture support/conflict passages for each claim.
4. Write one citation block per match using exact quotes and pinpoint citations.
5. Finish with a findings summary: alignments, discrepancies, and record gaps.
## Workflow
### 1. Anchor the Decision
- Treat the decision as the baseline source.
- Extract only claims that depend on testimony or the trial/hearing record.
- Keep each claim atomic (one proposition per claim) to avoid mixed citations.
### 2. Cross-Reference Transcripts
- Review each transcript session in chronological order.
- Tag each matched passage as:
- **`[ALIGNMENT]`** — transcript directly supports the decision's characterization
- **`[DISCREPANCY]`** — transcript contradicts, materially differs from, or fails to contain what the decision attributes to the record
### 3. Build Citation Blocks
Use this exact field order for every block:
- **`[ALIGNMENT]` or `[DISCREPANCY]`**
- **Document Name (Decision):** full decision filename/title
- **Location (Decision):** `Page X, Line Y` (or paragraph/section when line numbers do not exist)
- **Decision Quote:** exact verbatim excerpt
- **Document Name (Transcript):** full transcript filename/session label
- **Location (Transcript):** `Page X, Line Y`
- **Transcript Quote:** exact verbatim excerpt
- **Note:** optional 1-2 sentence explanation of why the pair is support or divergence
### 4. Summarize Findings
After all citation blocks, provide:
- **Primary Alignments** — highest-impact ways the record supports the decision
- **Primary Discrepancies** — highest-impact ways the record diverges from the decision
- **Record Gaps** — decision findings with no located supporting/opposing transcript passage
## Output Order
1. **Header block** listing all reviewed documents and transcript session labels
2. **Citation blocks** grouped by session, in session order
3. Within each session group: list **`[ALIGNMENT]`** blocks first, then **`[DISCREPANCY]`** blocks
4. **Large-corpus override:** if the corpus is very large or the user requests priority triage, list **`[DISCREPANCY]`** blocks first, then **`[ALIGNMENT]`** blocks
5. **Summary of Findings**
## Quality Controls
- Quote verbatim in both `Decision Quote` and `Transcript Quote`; do not paraphrase those fields.
- Never invent page/line references.
- If pinpoint location cannot be determined, write:
- `Page/Line unavailable — approximate location: <section or paragraph description>`
- Do not treat similar language as equivalent to an exact statement.
- If a decision paraphrases testimony, flag that in `Note`.
- Flag sessions with zero relevant passages; absence can be material.
- Mark uncertain citations or interpretations with `[VERIFY]`.
## Edge Cases
- **Multi-session transcripts**: treat each file as a distinct source and keep session label in every block.
- **Incomplete/redacted transcripts**: record the limitation under `Record Gaps`.
- **Privileged or sealed material**: do not reproduce protected excerpts; flag and advise handling under governing order.
- **Non-U.S. records**: adapt pinpoint format to the jurisdiction's transcript citation convention.