Files
mcp-server/spark_executor
ClaudeandClaude Fable 5 59cab3346e docs(get_connection): point to SHS, fetch_url, and the full field set
Old description was thin and only mentioned yarn_rm_url / auth_type
as the things to look up. That was written before we added
history_server_url and url_allowlist, and before fetch_url and
list_applications existed.

New description explicitly enumerates the connection fields an
agent is most likely to need (yarn_rm_url, history_server_url,
url_allowlist, auth_*, master/deploy_mode/spark_conf) and names
the tools that consume them (get_external_*, fetch_url,
list_applications, prepare_submit_job) so the LLM knows to call
get_connection first when it needs any of those.

Also adds a security note that the response includes secrets
(auth_password, auth_keytab) — the previous description didn't
warn about this.

Tests: 398 passed, no test changes (description text is not
asserted).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-09 16:26:54 +08:00
..
2026-06-24 11:19:17 +08:00