Update to Clio

{
  "description": "Search and read primary law in the vLex library, and get legal questions researched by Clio's agent — finding cases, statutes and commentary, reading a decision's text, checking whether an authority supports a proposition or a quotation is accurate, and checking the citations in a document the user supplies. Non-US work belongs here rather than with us-dockets.",
  "included_files": [],
  "name": "clio-research",
  "skill_md_contents": "---\nname: clio-research\ndescription: Search and read primary law in the vLex library, and get legal questions researched by Clio's agent — finding cases, statutes and commentary, reading a decision's text, checking whether an authority supports a proposition or a quotation is accurate, and checking the citations in a document the user supplies. Non-US work belongs here rather than with us-dockets.\n---\n\n# Clio Research\n\nTwo different costs, and choosing wrong is the most common mistake.\n\n| The user wants | Call | Cost |\n| --- | --- | --- |\n| **Documents** — does a case exist, find the statute, read this decision | `search_legal_documents`, then `get_legal_document` on a `vid` | one round trip, seconds |\n| **Reasoning** — a question answered, authorities weighed, an argument tested | `start_research` | minutes; the agent searches for itself |\n\n\"Is there a Spanish Supreme Court decision on shared custody and relocation?\" is a\nlookup. \"Can my client relocate with shared custody?\" is reasoning. When the user\nasks for both, do the lookup first — it is cheap and it grounds the question.\n\nWhen they asked only for reasoning, do **not** search first. The agent does its own\nsearching, so a scouting lookup repeats that work — and it renders a results card\nthat then sits beside the answer competing for attention, for a step the user never\nasked to see.\n\n## Searching\n\n`search_legal_documents(query=...)` takes the user's own terms. Narrow it only as\nfar as they did:\n\n- `jurisdiction` — country code (`ES`, `US`, `GB`). Pass it whenever they named a\n  country; never infer one from their location.\n- `sub_jurisdiction` — a state or region *within* that country, and only with a\n  `jurisdiction`. Note it is dropped for `posts` and `news`, which are tagged at\n  country level only, so a sub-jurisdiction there would filter every hit away.\n- `court` — a code from `search_courts`, not a name you composed. A guessed code\n  returns nothing rather than erroring, which reads as \"no such case\". Caselaw only.\n- `document_type` — `caselaw`, `legislation`, `posts`, `news`; leave blank for all.\n- `sort_by` — `most_recent` when recency is the point, otherwise leave it.\n\n`total_count` is how many **matched**, not how many came back. Say so: \"229 match,\nhere are the 10 most relevant\" is honest; presenting 10 as the whole answer is not.\n\n## Reading a document\n\n`get_legal_document(vid=...)` returns metadata plus `body_text`. An empty\n`body_text` means no format yielded text — report that rather than inferring the\ncontent from the title or headnotes.\n\nTreat `snippet` and `summary` from a search as extracts, never as the document's\nholding. And a hit is not authority: nothing in a search result says a case is\nstill good law.\n\n## Research that needs the agent\n\nFor reasoning, `start_research` with the question in the user's own terms, the\njurisdiction as a country code, and any material state or province as\n`sub_jurisdiction` (`texas`, `ontario`, `UK_SC`) as well as in the question\nitself. For drafting or review, include the represented party, purpose and\npriorities when known, and distinguish missing facts from assumptions.\n\nAsk Clio for supporting authorities and links, applicable statutory versions or\neffective dates, and contrary authority when it is material. Then handle the\n`TurnResult` exactly as **clio** describes.\n\nWhen the work is done, `show_cited_authorities` is the card for what it relied on.\nCall it once per conversation rather than per turn — it covers every turn.\n\nEvery authority you relied on also belongs in your reply, carrying its vLex link,\nbecause a host can fold a card inside a collapsed step where nobody reads it. When\nthe user asked for a list, `formatted` is that list already built. In a reasoned\nanswer, cite each authority where the reasoning uses it instead — a ten-row dump\npasted under a doctrinal answer reads worse than the answer alone.\n\n## Exporting an answer\n\nWhen Clio delivers a research memo, offer it as a file once: \"Do you want this as\na Word document or PDF?\" Offer the same when the user wants to share or file an\nanswer, or keep the whole exchange.\n\n- One answer: `export_conversation` with the reply's `message_id` from\n  `messages[]`. A memo exports in full with its authorities, as `docx` or `pdf`.\n  Any other reply exports as `docx` only.\n- The whole conversation: `export_conversation` without `message_id`, as `docx`.\n- A document Clio drafted is not an answer. Use `export_document` (see\n  **drafting**). If `export_conversation` returns a 400 that names a `file_key`,\n  call `export_document` with it.\n\nHandle the result as **drafting** describes for `export_document`: give the user\nthe `download_url` and tell them that anyone with the link can download the\nfile for seven days. Report a failure as a failure. Never fetch the link or\ninline the file. If the export is still rendering, check again with the same\narguments plus the returned `run_id`; a call without it starts another render.\n\n## Citations in a document the user supplies\n\nAttach the document as **clio** describes under **Files in**, then ask the agent\nto check its citations. Which call asks depends on what `attach_document` returned:\n\n- a `conversation_id` → `continue_conversation` on that conversation. The document\n  is already inside it; `start_research` would open a second conversation without it.\n- a `file_id` → `start_research(file_ids=[...])`, or\n  `continue_conversation(file_ids=[...])` when a run is already going.\n\nWhether one citation the user names exists is a lookup, per the table above. Every\nother check — a whole document's citations, or whether any citation supports its\nproposition, quotes accurately or remains good law — is reasoning.\n\nFor a whole document, ask the agent to list every citation it finds, each marked\nfound or not found in vLex. The list is the agent's own reading of the document,\nnot an exhaustive scan: it can miss citations or report errors the document does\nnot contain. When you can read the document yourself, check each claimed error\nagainst it before you report it. Tell the user the list may be incomplete, and\nnever present it as complete.\n"
}