Case File Proof Agent
Status: Live.
Your muse
Build a legal case agent like meta-models/legal-agent-workshop. It reads a fictional dispute file, drafts a checklist, proves each item with exact page quotes, asks me for evidence when it is stuck, and files a memo I sign off. Show every turn live.
Checks passed
The plan: Case File Proof Agent
Prove every checklist item from a fictional file before you sign.
Can build part of it
- Upload a fictional dispute PDF or paste fictional text
- Watch the live turn log as it drafts the proof checklist
- Answer evidence requests when it is stuck on an item
Won't do: This is a fresh stand-in inspired by meta-models/legal-agent-workshop
See the full plan
Who it's for
Law students and workshop learners practicing evidence-based review on fictional disputes.
How it works
- Upload a fictional dispute PDF or paste fictional text.
- Watch the live turn log as it drafts the proof checklist.
- Answer evidence requests when it is stuck on an item.
- Review each item with its exact page quotes.
- Sign off the final memo and download it.
Screens
- Case Setup. Upload the fictional file, set parties and rules, and start the live run.
- Live Case Run. Show every turn, the checklist, page quotes, and evidence questions in one feed.
- Memo Signoff. Review proved and stuck items, compare signatures, and sign the memo.
What goes in
- PDF: Fictional dispute file only, up to 30 pages. Real cases and personal data are not accepted.
- Text (optional): Short typed answers and pasted evidence when the agent asks for help. Kept in browser only.
What comes out
- Text: Live turn log showing each read, checklist draft, quote check, and user question.
- Text: Checklist with status per item plus exact page quotes for proved items.
- File: Final memo file to download after sign-off.
Models
Model Used for Why Muse Wonder Read fictional dispute file and run checklist, quoting, Q and A, and memo turns It reads long PDFs and text, keeps exact page quotes, asks for missing evidence, and drafts the final memo with a live turn log. SAM 3.1 Find signature regions on sign-off images for comparison It locates handwritten signature marks so the product can show the sign-off comparison next to the memo. Limits
- This is a fresh stand-in inspired by meta-models/legal-agent-workshop. It cannot clone or copy that repo.
- Fictional practice files only. Not legal advice and not for real disputes.
- Every proved item needs an exact page quote. Items without quotes stay stuck until you add evidence.
- Signature check is visual similarity only and does not verify identity.
- Large files are truncated to fit context, so very long exhibits may be missed.
From your inspo
No references were given. The note asks for a fresh legal case agent that reads a fictional file, proves each checklist item with quotes, asks for evidence when stuck, and files a sign-off memo with every turn shown live.
Takes
- The job of proving a fictional dispute file item by item
- The flow of draft checklist then page quotes then user evidence then signed memo
- The live turn by turn view of what the agent read and decided
Leaves out
- The name, code, text, and exact design of meta-models/legal-agent-workshop
- Any real cases, firms, logos, or copyrighted materials
Cost per run: about $2.33 (est.). About 14 Wonder turns with 180k input tokens at $4 per M = $0.72 plus 80k output tokens at $20 per M = $1.60 plus 2 SAM signature finds at $0.0025 each = $0.005, total about $2.33.
Recorded demo Done.
npm run verifyexits 0 with all 12 tests passing.It's live.