Benchmark report

Task verified

10m 57s elapsed

The saved attempt met the documented success criterion.

Experience score

99/ 100

Agent-Ready
  • SInstall
  • SAuth
  • SExecute
  • SDocs
  • ACurrent
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 S, Current ASInstallSAuthSExecuteSDocsACurrent
Agent profile · this run
Judge's note · what slowed the agent most
Probing the undocumented 5-decimal volume truncation (AssetPairs says lot_decimals=8) took several validate-only AddOrder calls and a code patch.

Where the time went

10:58 agent time
  • Starting0:01
  • Research0:23
  • Attempt7:50
  • Verify2:45
10m 57sTotal elapsed
6:40Time to first success
0:00Waited on the owner

Judge reason

Saved raw Kraken responses (real CF/X-Trace-Id headers) prove the full flow. Pre-check POST /0/private/OpenOrders returned {} and the run printed no ctrl0 order open (03_pre_openorders.json). AssetPairs gave tick_size 0.1, lot_decimals 8 (1e-8), ordermin 0.00006, costmin 0.5; Depth gave best bid 73686.60000 (02/04/05_*.json). Three separate POST /0/private/AddOrder calls (06/07/08) carried oflags=post, timeinforce=GTC, cl_ord_id ctrl0mo1/2/3 + shared suffix 1794455883, at 58949.2, 58212.4, 57475.5 (20/21/22% below bid = 58949.28/58212.414/57475.548, each rounded down within one 0.1 tick) with volumes 0.00017/0.00018/0.00018 -> notionals 10.02/10.48/10.35 USD (all >=10 and <11); txids OFZ4DJ-AD3KK-W3IVFL, ONNXC4-2OVOP-5RI3XU, OSJ63F-HFMEM-CNRJWX. 09_openorders_before_cancel.json shows all three open, vol_exec 0. A single POST /0/private/CancelOrderBatch with the three txids returned HTTP 200 {"error":[],"result":{"count":3}} (10). 11 and 13 show zero open orders on the pair; 12_queryorders_final.json shows all three status=canceled, vol_exec 0.00000000, cost 0.00000. Signing is hand-written stdlib HMAC-SHA512 (code/kraken_client.py) verified against the docs' worked example; no SDK/CLI/MCP used. Volume probes used validate=true only, so exactly three real orders existed. Credentials were requested as secret fields KRAKEN_API_KEY/KRAKEN_API_SECRET, read from files, sent only in API-Key/API-Sign headers, never logged; an in-run grep of every saved file for both values found no match and my own grep found only the docs' sample signature in 01_signing_selftest.txt.

Attempt record

  1. 0:53Answered from saved key: Kraken Spot REST API credentialsno wait
  2. —First success verified

Attempt artifacts saved.