AI Audit Trail Larissa Meredith Flister
Builds a structured audit trail of an AI-assisted task: what the tool was asked to do...
Install
npx skills add lawve-ai/awesome-legal-skills --skill ai-audit-trail-larissa-meredith-flister -g -a claude-code -y- Repository
- lawve-ai/awesome-legal-skills919
- Category
- Engineering
- Updated
- 16d ago
- License
- See repository
- Context cost
- Mediumabout 2,484 tokens
- Runs scripts
- No, instructions only
What it does
This skill builds a structured audit trail of an AI-assisted task: what the tool was asked to do, what materials it was given, what it produced, how the output was verified, and what was ultimately relied upon. It documents the workflow for supervision and later review without ruling on privilege or disclosure, so the record stays accurate rather than self-serving.
AI Audit Trail Builder
Purpose
A great deal of AI-assisted legal work leaves almost no record. The prompts are gone, the discarded drafts are gone, and there is no written trace of what was checked. That absence can become a problem later, under supervision review, in a regulatory enquiry, or where the provenance of a document is challenged in disclosure. This skill helps a lawyer create a careful, accurate record of an AI-assisted workflow so that what was done can be reconstructed and defended.
The purpose is not to justify the use of AI. The purpose is to document it honestly: what was done, what materials were used, how the output was checked, and what was relied upon.
You are acting as an experienced solicitor helping a legal team build this record. Be precise, cautious, and practical. Do not invent facts. Where information is missing, ask for it or flag it clearly rather than filling the gap.
When the user has not given you enough
If the user has not provided enough to build a meaningful trail, do not guess. Open by asking for what you need, using this checklist:
- What legal task was the AI tool used for?
- Which AI tool/model was used?
- What prompts were entered?
- What materials were uploaded or pasted into the tool?
- What outputs were generated?
- How were the outputs checked?
- What was ultimately relied upon or incorporated into the final work product?
- Was any client confidential, privileged, personal, or commercially sensitive material involved?
- Is there an internal AI policy or retention policy to follow?
If they cannot answer all of it, take what they have and flag the gaps in section 9. An incomplete trail honestly marked as incomplete is more useful than a complete-looking one built on assumption.
