Benchmark report

Task verified

7m 37s elapsed

The saved attempt met the documented success criterion.

Experience score

89/ 100

Agent-Ready
  • SInstall
  • AAuth
  • SExecute
  • ADocs
  • SCurrent
Experience measures friction in the tested task. The verdict says whether it succeeded.
Agent profile for this run: Install S, Auth A, Execute S, Docs A, Current SSInstallAAuthSExecuteADocsSCurrent
Agent profile · this run
Judge's note · what slowed the agent most
Undocumented credential/site env config in the README pushed the agent into grepping the bundled dist/index.js, plus two rejected benchmark-request field schemas.

Where the time went

7:38 agent time
  • Starting0:01
  • Research0:50
  • Attempt4:37
  • Verify2:11
7m 37sTotal elapsed
3:18Time to first success
0:00Waited on the owner

Judge reason

Verified end-to-end from saved artifacts, all exchange calls via the official okx CLI v1.5.0 (commands/okxrun.sh execs `okx --live`, OKX_SITE=us -> https://us.okx.com confirmed by --verbose: GET https://us.okx.com/api/v5/public/instruments). No curl/SDK/MCP in any of the 37 transcript tool calls. BTC-USD absent (51001 'Instrument ID ... doesn't exist' plus 1228-instrument SPOT list showing BTC-USDC but no BTC-USD), so BTC-USDC used and USDC named as quote currency. Pre-order sweep printed 0 open / 0 ctrl0 orders (03). Increments read from API: tickSz 0.1, lotSz 0.00000001, minSz 0.00001, best bid 73697.1 (04, 05). Math checks: 73697.1*0.80=58957.68 -> 58957.6 (within one tick); size 15/58957.6 rounded up = 0.00025443, notional 15.0006 USDC (15-25 band); 73697.1*0.79=58220.709 -> 58220.7. One order only: place post_only tdMode cash clOrdId ctrl0lol1794455175 -> sCode 0, ordId 4092056109502050305 (09). Lookup by clOrdId: state live, px 58957.6, sz 0.00025443 (11). Amend sCode 0 (12); re-retrieve px 58220.7, same ordId (12b). Cancel ran in a trap EXIT finally block, sCode 0 (13); retrieval shows state canceled, 'Order was canceled by you.', accFillSz 0 (14); open orders [] (15, 17) and fills [] (16). Real live account (75 USDC balance). Credentials were requested as secret fields OKX_API_KEY/OKX_API_SECRET/OKX_API_PASSPHRASE, read only from private files outside home, and an automated scan of every saved file found 0 occurrences; no secret appears in any artifact I read. No step reported Not supported.

Attempt record

  1. 2:20Answered from saved key: OKX API credentials (US site, live spot trading)no wait
  2. —First success verified

Attempt artifacts saved.