--- name: oscola-verify description: Verify citations in a legal draft — court level, reporter, neutral citation, pinpoints, and whether each authority was actually retrieved. Catches court-level misattribution, which is the error most likely to survive proofreading and destroy a paper in a viva. Use this on any paper, response paper, memo, or footnote set before submission, and whenever the user asks to check, verify, or clean up citations or footnotes. Do NOT use to generate new citations. --- # OSCOLA Verification The error this exists to catch is not a missing full stop. It is a High Court decision cited as Supreme Court authority, a case cited for a proposition it does not contain, or a pinpoint to a paragraph that says something else. Those survive proofreading because the citation *looks* right, and they are the ones that collapse under questioning. Formatting is checked too, but second. ## Order of checking Substance before form. A perfectly formatted citation to the wrong court is worse than a messy citation to the right one, because it is harder to spot. ### 1. Retrieval status Every authority gets one tag. This is the first thing to establish and the most important. - `[RETRIEVED]` — the text was fetched this session or supplied by the user. Name the source. - `[RECALLED — VERIFY]` — believed correct from background knowledge, not checked against the document. **This tag means the citation is not yet safe to submit.** - `[UNVERIFIABLE]` — searched for and not located. Report the count of each at the top of the output. A draft with eleven `[RECALLED — VERIFY]` citations is not ready, however well formatted. Never upgrade a tag without actually retrieving the document. Never invent a citation to fill a gap in the footnotes. ### 2. Court level and bench strength Check the court named on the face of the judgment against the court named in the footnote. This check must be performed individually for every case. It is tedious and it is the reason the skill exists. The specific failure to watch for: a High Court decision described or cited as though it were Supreme Court authority. This happens because the case is well known, because a secondary source described it loosely, or because the reporter abbreviation was misread. It is fatal in a paper that turns on precedential hierarchy. Where the argument depends on precedential weight, also record bench strength. In Indian constitutional and company law argument a two-judge bench cannot displace a settled three-judge position, and an argument built without noticing the numbers will not survive a viva. ### 3. Proposition match For every citation supporting a substantive claim, confirm the cited passage actually contains the proposition. Distinguish: - **Ratio** — the reasoning necessary to the decision, binding. - **Obiter** — reasoning not necessary to the decision, persuasive only. - **Argument recorded** — counsel's submission summarised in the judgment. This is not the court's view and must never be cited as though it were. A frequent and serious error. - **Dissent** — must be identified as such. Where a case is cited for a proposition that appears only in a dissent, or only in the recitation of argument, flag it plainly. ### 4. Currency Has the authority been overruled, doubted, distinguished into irrelevance, or referred to a larger bench? Does the statutory provision cited still exist in that form? Judgments predating 1 July 2024 cite the Indian Penal Code, Code of Criminal Procedure and Indian Evidence Act rather than their replacements; where such a judgment is relied on, the mapping should be stated rather than silently performed. ### 5. Format Only now. Read `references/oscola-india.md` for the conventions, including where Indian practice departs from OSCOLA as written for England and Wales. Check consistency before correctness where the two conflict: a draft that uses one convention throughout is better than one that mixes two correct conventions. Where the user's institution has a house style, that governs — OSCOLA has no Indian-specific edition, and law schools differ in how they adapt it. Ask which convention the course requires if it is not evident from the draft, rather than imposing one. ## Output ``` ## Retrieval summary Retrieved: n · Recalled, unverified: n · Unverifiable: n ## Substantive problems [Court-level errors, proposition mismatches, obiter cited as ratio, argument cited as holding, overruled authority. One entry per problem, with the footnote number, what the draft says, and what the source says.] ## Currency flags [Repealed provisions, superseded statutes, pending references.] ## Format corrections [Grouped by type rather than listed footnote by footnote, so the user learns the pattern. Give the rule, then the instances.] ## Still to verify [Every [RECALLED — VERIFY] citation, listed, with the specific search or database the user should use to close it.] ``` ## Constraints Do not generate citations. If the draft asserts a proposition without support, say that support is missing and describe what kind of authority would supply it. Do not supply a case name that seems likely to fit. Do not silently correct a citation to a different case because the cited one does not support the claim. Report the mismatch and let the author decide whether the claim or the citation is wrong. Those are different repairs with different consequences for the argument. Where uncertain about a formatting convention, say so rather than asserting a rule. An invented rule applied consistently across a paper produces a consistent error, which is worse than an inconsistency the marker would have overlooked.