Research & verification

Every legal proposition in the skills is verified before it ships: statutory text from the Missouri Revisor rather than aggregators, every case checked through CourtListener for existence and reporter citation, pin cites to the subsection, date stamps on volatile figures, and contingent holdings marked contingent. The full method — including the eleven errors made while applying it, published because the errors are the argument for the protocol — is at /verification-protocol/.

mo-authority.md is the shared canonical authority file for the whole library: every proposition the skills assert, pin-cited and dated, with a quarterly review trigger for figures that adjust annually. Each skill also ships a stage-specific authority.md. Propositions that could not be verified are marked ⛔ OPEN and the skills will not assert them.

• Verified authority — a proposition supported by primary-source text (Revisor) or a CourtListener-verified opinion, pin-cited to the subsection or page, with a verification date.
• Synthetic example — demonstration facts invented for illustration and labeled as such. No sample on this site comes from a real matter.
• Open question — a legal issue with no controlling authority (or contested authority), marked ⛔ OPEN in the authority file; the skills present it as live rather than settled.
• Unverified proposition — a statement that has not passed the protocol. Unverified propositions do not ship in skill instructions.

The v0.2.0 release corrected four substantive defects found by applying the protocol to the library's own v0.1.0 text, including: demands rebuilt to the § 408.040.3 / § 537.058 statutory union (the old 30-day default silently forfeited prejudgment interest); the billed-amount subtotal replaced with billed/adjusted/paid/outstanding after § 490.715.5; hospital-lien math corrected to § 430.225.3; and the wrongful-death watch date computed from the date of death rather than the crash. The full history is in the changelog and the RSS feed.

Planned, not yet published:

• A synthetic-matter benchmark set (labeled synthetic) for measuring skill output against the authority file.
• Structured comparisons of skill-guided output versus generic prompting on the same synthetic facts.
• Quarterly re-verification reports for the date-stamped figures in mo-authority.md.

Nothing on this page has undergone external peer review, and no such claim is made. If you run a study using these materials, we would like to link it here — see the contact route below.

Found an error in a proposition, citation, or deadline? Open an issue with the authority-correction template — it asks for the exact proposition, the proposed source with URL and pin cite, an access date, and whether the error affects skill output or only explanatory text. General contact runs through GitHub issues.