No user-submitted related servers found.
Scored across 20 tools
Several tools have genuinely overlapping territory: tandem_start vs tandem_continue (both begin task/goal execution), tandem_review vs tandem_review_run (both produce review bundles/runs), and tandem_result vs tandem_reply vs tandem_wait (all mediate interacting with a running task). The very detailed descriptions help an agent differentiate them, but the boundaries are subtle enough that misselection is plausible.
Every tool uses the same tandem_ prefix with snake_case verb_noun or noun forms (tandem_start, tandem_result, tandem_reply, tandem_cancel, tandem_list). Minor noun-only names like tandem_channel or tandem_scope remain fully readable and predictable within the established convention.
Twenty tools is on the heavy side for the rubric's 16-25 'borderline heavy' band. The domain is genuinely complex (task orchestration, review lifecycle, context sharing, channels, diagnostics), so most tools earn a place, but the surface is larger than an agent can comfortably hold.
The lifecycle is well covered: starting, continuing, waiting, reading results, replying, cancelling, listing, plus review, findings, artifacts, context publication/export/import, channels, receipts, and diagnosis. Few obvious dead ends exist, though some operations (e.g. task mutation nuances) lean on one large multi-action tool rather than dedicated endpoints.