The org's data connections, names + scopes only (never material)

Your org's stored data connections, BY NAME ONLY - the access they grant is exercised only by the platform and is never shown to you or anyone: there is nothing here to paste anywhere. Returns [{name, kind, verbs, scope, auth, disabled, last_used_at}], e.g. {name: "acme-corpus", kind: "s3-keys", verbs: ["read"], scope: "s3://acme-ml/corpus-v3/*", auth: "token"} (auth "oauth" = a browser grant, no stored secret at all; one still pending its browser step shows pending_authorization: true and can't import yet). Use a name with import_data to land private data on a node without ever holding a credential. A kind "registry" row is different: it is a private Docker registry, exercised when a JOB pulls an image inside its scope, so you never name it anywhere - just pass the image to submit_job(image:) and the platform authenticates the pull. Empty, or missing the source you need? A human org admin adds one from their own terminal or CI - autor connection create (install: curl -fsSL https://autoresearch.sfcompute.com/cli.sh | bash, then autor login) - a one-time step, ~2 minutes, good for every future task. Ask them to do exactly that; NEVER ask anyone to paste a token, key, or password into this chat (transcripts are recorded and outlive the task), and connection mutations are not callable with this connection's grant.