runnerID by Torchrunner

Most tools target distinct resource+action pairs, and descriptions explicitly cross-reference alternatives (e.g., 'Prefer fill_code when you have a cdp_url'). However, the email-code family (wait_for_email_code, get_email_code, fill_email_code) and the authenticator pair overlap enough that an agent must read carefully to choose correctly, which is why it falls short of a perfect score.

Every tool follows a consistent snake_case verb_noun pattern (get_*, save_*, fill_*, wait_for_*, cancel_run, open_magic_link, handoff_task). The verb semantics are used consistently: fill_* types into the page, get_* returns data, save_* persists secrets.

18 tools is at the upper bound for this domain but each covers a meaningful credential/browser workflow step. One or two (notably wait_for_email_code alongside fill_email_code) feel nearly redundant, keeping it just below a 5.

The surface covers runs (handoff/get/cancel), sessions (save/restore), logins, authenticator seeds, email codes, approvals, events, and magic links—a coherent lifecycle. Gaps exist around lifecycle management: there is no delete/revoke for stored logins, sessions, or authenticator seeds, and no list_identities, only get_identity for a single known identity.