AI is fast and fluent, and fluency is not accuracy. A confident, well-formatted answer can still contain a fabricated case, a wrong number, or a diagnosis that isn't in the record. Every skill here is built around that reality — and so should your habits be.
Treat every AI output as a lead, never a source. The signature on the demand is yours, not the model's.
01
Output is a lead, not a source
Every figure, citation, and diagnosis routes back to the underlying record or authority before it leaves your office.
02
Demonstrate the failure once
Watch a confident answer collapse under verification, and the verification habit sticks for good.
03
Guard what you paste
Be deliberate about which client facts enter any tool. Confidentiality is a decision you make on every prompt.
04
The signature is yours
The skill drafts and flags its assumptions. You verify, you decide, you sign. That step is never delegated.
Two duties to keep front of mind. Confidentiality — be deliberate about what client information goes into any tool. Candor and competence — you remain responsible for every fact and citation that leaves your office, exactly as if you'd written it yourself.
We hold the skills to the same standard. As of v0.2.0 every legal proposition in this library is pin-cited to the subsection and dated in
mo-authority.md, verified against the Revisor and CourtListener rather than against memory. Propositions we could not verify are marked
OPEN and the skills will not assert them. The method, the four substantive defects v0.2.0 fixed, and the eleven errors we made finding them are all published at
/verification-protocol. A library that tells you to check its work should show you its own.
01 · What the skills refuse to do
- Invent facts. Missing providers, dates, bills, addresses, and claim numbers come back as placeholders or open questions — never as plausible filler.
- Assert unverified law. Propositions marked
OPEN in the authority file are not stated by the skills, in either direction.
- Quote figures from memory. Amounts that adjust annually (the sovereign-immunity cap, statutory fee schedules) must be pulled and date-stamped, never recalled.
- Set terms on their own. Time-limited and policy-limits demand terms are flagged for the attorney rather than chosen by the skill.
02 · How to read the flags
[CONFIRM WITH ATTORNEY] — a computed date or deadline presented as a watch item, never as settled advice.[UNVERIFIED] — record content that was illegible, ambiguous, or absent.[ASSUMPTION] — an inference the draft depends on; resolve or escalate it before the draft goes out.[ESTIMATE — basis: …] — a number that is not a hard figure, with its basis stated.- "Items for attorney review" — the closing block on a demand draft. It is a work list, not a formality.
03 · A verification habit that fits in a workday
- Read the flags first, before the prose. They tell you where the draft is load-bearing.
- Trace every dollar figure to the bill or record it came from.
- Open every statute cited and read the subsection — not a summary site.
- Confirm every case exists, is cited to the right reporter, and has no later history that changes it.
- Ask what the draft did not say. Silence is where a chronology gap or a missing party hides.
The verification protocol is the long-form version of this list, applied to the library's own authority file. The CLE materials turn it into a teachable exercise.
These skills are drafting aids for licensed attorneys and their staff. They are not legal advice, they do not create an attorney-client relationship, and they do not replace attorney review.