Benchmark report

Task verified

5m 42s elapsed

The saved attempt met the documented success criterion.

Experience score

96/ 100

Agent-Ready
  • SInstall
  • SAuth
  • SExecute
  • ADocs
  • SCurrent
Experience measures friction in the tested task. The verdict says whether it succeeded.
Agent profile for this run: Install S, Auth S, Execute S, Docs A, Current SSInstallSAuthSExecuteADocsSCurrent
Agent profile · this run
Judge's note · what slowed the agent most
One wasted batch-cancel call: product_type=SPOT, valid on other coinbase tools, returned HTTP 400 'unknown field product_type'; resend without it succeeded.

Where the time went

5:42 agent time
  • Research0:37
  • Attempt3:08
  • Verify1:57
5m 42sTotal elapsed
1:52Time to first success
0:00Waited on the owner

Judge reason

Transcript MCP tool results match every saved artifact. Live env confirmed (coinbase_env -> 'Active: live'). Pre-check coinbase_orders_list on BTC-USD returned empty (no ctrl0 order to cancel). coinbase_products_get printed quote_increment 0.01, base_increment 0.00000001, base_min_size 0.00000001/quote_min_size 1; best bid 73481.5 from coinbase_products_best_bid_ask. Three separate coinbase_orders_create calls placed post-only GTC limit buys at 58785.20/58050.38/57315.57 (I recomputed 20/21/22% below 73481.5 rounded down to 0.01 — exact match) with sizes 0.00017012/0.00017227/0.00017448 (10 USD / price rounded up to 1e-8; notionals 10.0005, 10.0003, 10.0004 — all >=10 and <11), client order IDs ctrl0mo1/2/3 + shared suffix 1794469313, order_ids 4f74419d..., ec0aff1e..., c110c5a0.... The OPEN listing before cancel showed all three. One coinbase_orders_cancel request with the three order_ids returned results success:true for each (a prior attempt including product_type was rejected HTTP 400 'unknown field product_type' and never reached the exchange, so the resend created no double-cancel). Final OPEN listing empty; coinbase_orders_get on each shows CANCELLED, filled_size 0, number_of_fills 0. No curl/HTTP/CLI call touched the exchange; no benchmark-request credential call; grep for key/secret material in /workspace/home found nothing.

Attempt record

  1. —First success verified

Attempt artifacts saved.