Update to MORVS

{
  "description": "This skill should be used when a user asks to find a MORVS or Voidly public dataset, identify a keyless JSON endpoint, check dataset dates and coverage, or understand source attribution and reuse terms.",
  "included_files": [],
  "name": "dataset-finder",
  "skill_md_contents": "---\nname: dataset-finder\ndescription: This skill should be used when a user asks to find a MORVS or Voidly public dataset, identify a keyless JSON endpoint, check dataset dates and coverage, or understand source attribution and reuse terms.\n---\n\n# Dataset finder\n\nThe public [MORVS and Voidly dataset catalog](https://morvs.ai/voidly/datasets.json) is a static, keyless JSON document. Fetch it when answering; do not rely on a remembered entry or count. It is distinct from MORVS's federal data hub and its MCP tools.\n\n## Workflow\n\n1. Fetch the catalog through an available read-only HTTPS tool, such as Claude Code WebFetch or Codex web.run. Search entries by topic, name, ID, and description. Do not invent a catalog search MCP tool; none appears in the 2026-10-09 live 73-tool list.\n2. For each relevant entry, record the landing page, JSON endpoint, extra endpoints, description, counts, build date or observed-at time, and stated license or rights. Follow the entry's endpoint and source/provenance links before claiming current coverage. A catalog count is a dated observation, not a live promise.\n3. Read license and rights at both catalog and entry level. For entries derived from third-party material, follow upstream terms; a compilation license does not automatically cover raw source data. State attribution and commercial-reuse limits only when verified at the linked terms.\n4. For a federal data hub source found through an entity, call `resolve` with a ticker, CIK, or name to inspect the available-in map, then call `dataset_freshness` with `{}` for source-specific cadence, last successful observation, and indexed-content watermark. Do not apply that freshness result to the static catalog entries unless the dataset is actually represented in the returned row.\n5. Return only institution-level or aggregate facts. Avoid PII, unsupported accusations, or treating data absence as evidence of no activity.\n\n| Dataset | Best use | Landing / JSON endpoint | Data date and coverage | Rights / upstream terms | Checked on |\n|---|---|---|---|---|---|\n| Catalog name | Plain-language use | Both clickable links | Explicit date semantics and gaps | Exact catalog or source terms | Access date |\n\n## Worked examples\n\n**Banned-books data.** For “Find a reusable banned-books JSON source”, use WebFetch or web.run to fetch [the catalog](https://morvs.ai/voidly/datasets.json), select its Verboten entry, and open its listed landing page and endpoint. Read the entry's license and source links. Report the current endpoint, its date/coverage, and the stated terms with citations; do not copy a remembered catalog count as live coverage. No MORVS MCP catalog tool is needed.\n\n**Company-linked federal dataset.** For “Which MORVS federal datasets mention Pfizer?”, call `resolve` with `{\"ticker\":\"PFE\"}`. Use its available-in map and links to choose datasets, then call `dataset_freshness` with `{}` and match returned dataset rows. If the user also wants a static Voidly dataset, fetch [the separate catalog](https://morvs.ai/voidly/datasets.json) and check that entry's own endpoint and rights. Keep the two coverage systems separate.\n"
}