@tangle-network/agent-bench 0.3.6 → 0.3.8

package/src/quant-arena/fixtures/demo-campaign/cost-ledger.jsonl

@@ -0,0 +1,16 @@
+{"version":1,"record":{"status":"pending","callId":"f0af6aca-8e22-4264-91d0-81f1a391e345","channel":"driver","phase":"search.proposal","actor":"proposer-shot:default-author","model":"sonnet","timestamp":1784753519601,"tags":{"candidateId":"cand-001-default-author"}}}
+{"version":1,"record":{"status":"settled","callId":"f0af6aca-8e22-4264-91d0-81f1a391e345","channel":"driver","phase":"search.proposal","actor":"proposer-shot:default-author","model":"sonnet","timestamp":1784753519601,"inputTokens":2,"outputTokens":12384,"cachedTokens":14844,"costUsd":0.36238919999999997,"costUnknown":false,"usageUnknown":false,"actualCostUsd":0.36238919999999997,"tags":{"candidateId":"cand-001-default-author"}}}
+{"version":1,"record":{"status":"pending","callId":"cc0c7780-83e5-4f6b-8cff-42ce8fcee086","channel":"driver","phase":"audit.leak","actor":"leak-auditor:claude","model":"haiku","timestamp":1784753648815,"tags":{"candidateId":"cand-001-default-author","attempt":"0"}}}
+{"version":1,"record":{"status":"settled","callId":"cc0c7780-83e5-4f6b-8cff-42ce8fcee086","channel":"driver","phase":"audit.leak","actor":"leak-auditor:claude","model":"haiku","timestamp":1784753648815,"inputTokens":10,"outputTokens":5672,"cachedTokens":10942,"costUsd":0.0626702,"costUnknown":false,"usageUnknown":false,"actualCostUsd":0.0626702,"tags":{"candidateId":"cand-001-default-author","attempt":"0"}}}
+{"version":1,"record":{"status":"pending","callId":"15f76357-12e9-4164-a575-609ea19d6caf","channel":"driver","phase":"search.proposal","actor":"proposer-shot:default-author","model":"sonnet","timestamp":1784753707786,"tags":{"candidateId":"cand-002-default-author"}}}
+{"version":1,"record":{"status":"settled","callId":"15f76357-12e9-4164-a575-609ea19d6caf","channel":"driver","phase":"search.proposal","actor":"proposer-shot:default-author","model":"sonnet","timestamp":1784753707786,"inputTokens":2,"outputTokens":14095,"cachedTokens":20285,"costUsd":0.3580005,"costUnknown":false,"usageUnknown":false,"actualCostUsd":0.3580005,"tags":{"candidateId":"cand-002-default-author"}}}
+{"version":1,"record":{"status":"pending","callId":"544ff960-2698-4b41-a86e-4a43da853e15","channel":"driver","phase":"audit.leak","actor":"leak-auditor:claude","model":"haiku","timestamp":1784753850937,"tags":{"candidateId":"cand-002-default-author","attempt":"0"}}}
+{"version":1,"record":{"status":"settled","callId":"544ff960-2698-4b41-a86e-4a43da853e15","channel":"driver","phase":"audit.leak","actor":"leak-auditor:claude","model":"haiku","timestamp":1784753850937,"inputTokens":10,"outputTokens":3333,"cachedTokens":14989,"costUsd":0.0432539,"costUnknown":false,"usageUnknown":false,"actualCostUsd":0.0432539,"tags":{"candidateId":"cand-002-default-author","attempt":"0"}}}
+{"version":1,"record":{"status":"pending","callId":"7c240191-60d7-4322-868f-86dee8a8356b","channel":"driver","phase":"search.proposal","actor":"proposer-shot:quant-researcher","model":"sonnet","timestamp":1784753888358,"tags":{"candidateId":"cand-003-quant-researcher"}}}
+{"version":1,"record":{"status":"settled","callId":"7c240191-60d7-4322-868f-86dee8a8356b","channel":"driver","phase":"search.proposal","actor":"proposer-shot:quant-researcher","model":"sonnet","timestamp":1784753888358,"inputTokens":2,"outputTokens":16166,"cachedTokens":14853,"costUsd":0.4205679,"costUnknown":false,"usageUnknown":false,"actualCostUsd":0.4205679,"tags":{"candidateId":"cand-003-quant-researcher"}}}
+{"version":1,"record":{"status":"pending","callId":"5492fe01-633d-4513-800b-194834546a77","channel":"driver","phase":"audit.leak","actor":"leak-auditor:claude","model":"haiku","timestamp":1784754065621,"tags":{"candidateId":"cand-003-quant-researcher","attempt":"0"}}}
+{"version":1,"record":{"status":"settled","callId":"5492fe01-633d-4513-800b-194834546a77","channel":"driver","phase":"audit.leak","actor":"leak-auditor:claude","model":"haiku","timestamp":1784754065621,"inputTokens":10,"outputTokens":10662,"cachedTokens":14989,"costUsd":0.0799569,"costUnknown":false,"usageUnknown":false,"actualCostUsd":0.0799569,"tags":{"candidateId":"cand-003-quant-researcher","attempt":"0"}}}
+{"version":1,"record":{"status":"pending","callId":"d5f20214-c3f3-472a-ada6-130aeec8e76b","channel":"driver","phase":"search.proposal","actor":"proposer-shot:quant-researcher","model":"sonnet","timestamp":1784754170071,"tags":{"candidateId":"cand-004-quant-researcher"}}}
+{"version":1,"record":{"status":"settled","callId":"d5f20214-c3f3-472a-ada6-130aeec8e76b","channel":"driver","phase":"search.proposal","actor":"proposer-shot:quant-researcher","model":"sonnet","timestamp":1784754170071,"inputTokens":2,"outputTokens":7930,"cachedTokens":20449,"costUsd":0.2651307,"costUnknown":false,"usageUnknown":false,"actualCostUsd":0.2651307,"tags":{"candidateId":"cand-004-quant-researcher"}}}
+{"version":1,"record":{"status":"pending","callId":"a9f6a64b-2c17-4100-b1a1-cca2ce406ab7","channel":"driver","phase":"audit.leak","actor":"leak-auditor:claude","model":"haiku","timestamp":1784754251978,"tags":{"candidateId":"cand-004-quant-researcher","attempt":"0"}}}
+{"version":1,"record":{"status":"settled","callId":"a9f6a64b-2c17-4100-b1a1-cca2ce406ab7","channel":"driver","phase":"audit.leak","actor":"leak-auditor:claude","model":"haiku","timestamp":1784754251978,"inputTokens":10,"outputTokens":6020,"cachedTokens":14989,"costUsd":0.0571829,"costUnknown":false,"usageUnknown":false,"actualCostUsd":0.0571829,"tags":{"candidateId":"cand-004-quant-researcher","attempt":"0"}}}

package/src/quant-arena/fixtures/demo-campaign/notebook.jsonl

@@ -0,0 +1,5 @@
+{"schema":"quant-arena.baselines.v1","at":"2026-07-22T20:51:59.593Z","seed":20260722,"costBps":10,"slippageBps":5,"windows":[{"start":199,"end":703,"startDate":"2017-04-06","endDate":"2019-03-12"},{"start":468,"end":972,"startDate":"2018-04-18","endDate":"2020-03-23"},{"start":546,"end":1050,"startDate":"2018-08-06","endDate":"2020-07-09"},{"start":594,"end":1098,"startDate":"2018-10-11","endDate":"2020-09-15"},{"start":680,"end":1184,"startDate":"2019-02-08","endDate":"2021-01-13"},{"start":1014,"end":1518,"startDate":"2020-05-21","endDate":"2022-04-26"},{"start":1057,"end":1561,"startDate":"2020-07-21","endDate":"2022-06-24"},{"start":1449,"end":1953,"startDate":"2022-01-20","endDate":"2023-12-26"}],"perWindowSharpe":{"buy-hold-index":[1.2094835481971722,0.2675576296109006,0.022309846317549994,-0.0258536811697932,0.3077114359626541,0.17304649923812107,0.0849232983176449,0.7851782377956339],"equal-weight":[1.1460678820228547,0.6061551930118717,0.2850058348026951,0.17908115506948746,0.660766081694157,0.5185595955384733,0.5223261328904791,1.214848475647775],"sma-crossover":[1.1152950659005165,0.0909886478242654,-0.05356223388832718,-0.014348351421960373,0.35996356736295215,0.2507467148916318,0.20379525595898224,1.2082647202257235]},"bestPerWindow":[1.2094835481971722,0.6061551930118717,0.2850058348026951,0.17908115506948746,0.660766081694157,0.5185595955384733,0.5223261328904791,1.214848475647775],"fullSample":{"buy-hold-index":{"start":0,"end":2086,"days":2086,"totalReturn":1.1730593785405552,"maxDrawdown":0.28396402140244215,"sharpe":0.5895302710354368,"tradeCount":1,"turnover":1},"equal-weight":{"start":0,"end":2086,"days":2086,"totalReturn":2.756912129075394,"maxDrawdown":0.26001662984291696,"sharpe":0.8405076018418971,"tradeCount":1100,"turnover":6.9338160520867085},"sma-crossover":{"start":0,"end":2086,"days":2086,"totalReturn":1.777456044807273,"maxDrawdown":0.3575492066705734,"sharpe":0.6592535650279422,"tradeCount":1789,"turnover":111.15341469281668}}}
+{"schema":"quant-arena.candidate.v1","at":"2026-07-22T20:51:59.600Z","candidateId":"cand-001-default-author","proposer":"default-author","authorModel":"sonnet","strategyPath":"/tmp/claude-1000/-home-drew-code-supervisor-lab/f06fd156-042a-4ef9-bd88-f2ec7f52b90c/scratchpad/qa-demo/strategies/cand-001-default-author/strategy.ts","sha256":"sha256:942976dc88981ab88b05602e023769c3e9a730133dec85aefdf49c7be7d41d44","authoringCostUsd":0.36238919999999997,"nTried":1,"leakAudit":{"truncation":{"clean":true,"cutoffs":[121,514,906,1299,1691,2084],"divergence":null},"llm":{"verdict":"clean","evidence":""}},"eval":{"perWindow":[{"start":199,"end":703,"startDate":"2017-04-06","endDate":"2019-03-12","sharpe":0.4152508658831424,"bestBaselineSharpe":1.2094835481971722,"excess":-0.7942326823140298},{"start":468,"end":972,"startDate":"2018-04-18","endDate":"2020-03-23","sharpe":-0.1221685157179545,"bestBaselineSharpe":0.6061551930118717,"excess":-0.7283237087298262},{"start":546,"end":1050,"startDate":"2018-08-06","endDate":"2020-07-09","sharpe":-0.3255205884630931,"bestBaselineSharpe":0.2850058348026951,"excess":-0.6105264232657882},{"start":594,"end":1098,"startDate":"2018-10-11","endDate":"2020-09-15","sharpe":-0.24745501555808566,"bestBaselineSharpe":0.17908115506948746,"excess":-0.4265361706275731},{"start":680,"end":1184,"startDate":"2019-02-08","endDate":"2021-01-13","sharpe":0.22164394268361348,"bestBaselineSharpe":0.660766081694157,"excess":-0.4391221390105435},{"start":1014,"end":1518,"startDate":"2020-05-21","endDate":"2022-04-26","sharpe":0.22882389795977512,"bestBaselineSharpe":0.5185595955384733,"excess":-0.2897356975786982},{"start":1057,"end":1561,"startDate":"2020-07-21","endDate":"2022-06-24","sharpe":0.23448567716788063,"bestBaselineSharpe":0.5223261328904791,"excess":-0.28784045572259853},{"start":1449,"end":1953,"startDate":"2022-01-20","endDate":"2023-12-26","sharpe":1.397778520983859,"bestBaselineSharpe":1.214848475647775,"excess":0.1829300453360838}],"meanExcessSharpe":-0.4241734039891216,"wins":1,"requiredWins":6,"threshold":0.1},"inSampleFull":{"start":0,"end":2086,"days":2086,"totalReturn":0.409063349034412,"maxDrawdown":0.14736423468316553,"sharpe":0.5043406767893703,"tradeCount":1703,"turnover":59.702106319096856},"verdict":"rejected-no-edge","reasons":["consistency: beat the best baseline in only 1/8 windows (need 6)","multiplicity: mean excess Sharpe -0.424 < required 0.100 (bar after 1 tried candidate)"]}
+{"schema":"quant-arena.candidate.v1","at":"2026-07-22T20:55:07.785Z","candidateId":"cand-002-default-author","proposer":"default-author","authorModel":"sonnet","strategyPath":"/tmp/claude-1000/-home-drew-code-supervisor-lab/f06fd156-042a-4ef9-bd88-f2ec7f52b90c/scratchpad/qa-demo/strategies/cand-002-default-author/strategy.ts","sha256":"sha256:8e81dd7939b38e9af3323fab63f7a9be18961bee660c6a43736ff0a4086683d9","authoringCostUsd":0.3580005,"nTried":2,"leakAudit":{"truncation":{"clean":true,"cutoffs":[121,514,906,1299,1691,2084],"divergence":null},"llm":{"verdict":"clean","evidence":""}},"eval":{"perWindow":[{"start":199,"end":703,"startDate":"2017-04-06","endDate":"2019-03-12","sharpe":0.4361239890211772,"bestBaselineSharpe":1.2094835481971722,"excess":-0.773359559175995},{"start":468,"end":972,"startDate":"2018-04-18","endDate":"2020-03-23","sharpe":0.04317392068219127,"bestBaselineSharpe":0.6061551930118717,"excess":-0.5629812723296804},{"start":546,"end":1050,"startDate":"2018-08-06","endDate":"2020-07-09","sharpe":-0.37020834846200457,"bestBaselineSharpe":0.2850058348026951,"excess":-0.6552141832646996},{"start":594,"end":1098,"startDate":"2018-10-11","endDate":"2020-09-15","sharpe":-0.5064161668353256,"bestBaselineSharpe":0.17908115506948746,"excess":-0.685497321904813},{"start":680,"end":1184,"startDate":"2019-02-08","endDate":"2021-01-13","sharpe":0.2622191188983631,"bestBaselineSharpe":0.660766081694157,"excess":-0.3985469627957939},{"start":1014,"end":1518,"startDate":"2020-05-21","endDate":"2022-04-26","sharpe":0.20867457049898017,"bestBaselineSharpe":0.5185595955384733,"excess":-0.3098850250394931},{"start":1057,"end":1561,"startDate":"2020-07-21","endDate":"2022-06-24","sharpe":0.403492143125228,"bestBaselineSharpe":0.5223261328904791,"excess":-0.11883398976525111},{"start":1449,"end":1953,"startDate":"2022-01-20","endDate":"2023-12-26","sharpe":0.8154427981443754,"bestBaselineSharpe":1.214848475647775,"excess":-0.3994056775033997}],"meanExcessSharpe":-0.4879654989723907,"wins":0,"requiredWins":6,"threshold":0.2766115033773212},"inSampleFull":{"start":0,"end":2086,"days":2086,"totalReturn":0.193975994870764,"maxDrawdown":0.2551410492136035,"sharpe":0.25557327907858696,"tradeCount":3862,"turnover":170.38253887993517},"verdict":"rejected-no-edge","reasons":["consistency: beat the best baseline in only 0/8 windows (need 6)","multiplicity: mean excess Sharpe -0.488 < required 0.277 (bar after 2 tried candidates)"]}
+{"schema":"quant-arena.candidate.v1","at":"2026-07-22T20:58:08.358Z","candidateId":"cand-003-quant-researcher","proposer":"quant-researcher","authorModel":"sonnet","strategyPath":"/tmp/claude-1000/-home-drew-code-supervisor-lab/f06fd156-042a-4ef9-bd88-f2ec7f52b90c/scratchpad/qa-demo/strategies/cand-003-quant-researcher/strategy.ts","sha256":"sha256:560aa30401092c749a42bb51329afa56929390994c957f77486d1b4bc4203804","authoringCostUsd":0.4205679,"nTried":3,"leakAudit":{"truncation":{"clean":true,"cutoffs":[121,514,906,1299,1691,2084],"divergence":null},"llm":{"verdict":"clean","evidence":""}},"eval":{"perWindow":[{"start":199,"end":703,"startDate":"2017-04-06","endDate":"2019-03-12","sharpe":0.4473951213548188,"bestBaselineSharpe":1.2094835481971722,"excess":-0.7620884268423533},{"start":468,"end":972,"startDate":"2018-04-18","endDate":"2020-03-23","sharpe":0.2641934383071636,"bestBaselineSharpe":0.6061551930118717,"excess":-0.34196175470470813},{"start":546,"end":1050,"startDate":"2018-08-06","endDate":"2020-07-09","sharpe":0.017374599330440648,"bestBaselineSharpe":0.2850058348026951,"excess":-0.2676312354722544},{"start":594,"end":1098,"startDate":"2018-10-11","endDate":"2020-09-15","sharpe":-0.2012077443814074,"bestBaselineSharpe":0.17908115506948746,"excess":-0.38028889945089484},{"start":680,"end":1184,"startDate":"2019-02-08","endDate":"2021-01-13","sharpe":0.47142122049408347,"bestBaselineSharpe":0.660766081694157,"excess":-0.1893448612000735},{"start":1014,"end":1518,"startDate":"2020-05-21","endDate":"2022-04-26","sharpe":0.21206398259244125,"bestBaselineSharpe":0.5185595955384733,"excess":-0.30649561294603206},{"start":1057,"end":1561,"startDate":"2020-07-21","endDate":"2022-06-24","sharpe":0.10190929084441298,"bestBaselineSharpe":0.5223261328904791,"excess":-0.42041684204606616},{"start":1449,"end":1953,"startDate":"2022-01-20","endDate":"2023-12-26","sharpe":0.5759685014944753,"bestBaselineSharpe":1.214848475647775,"excess":-0.6388799741532998}],"meanExcessSharpe":-0.4133884508519602,"wins":0,"requiredWins":6,"threshold":0.3223455711051266},"inSampleFull":{"start":0,"end":2086,"days":2086,"totalReturn":0.26253273244807596,"maxDrawdown":0.1824773874551917,"sharpe":0.33772524926578196,"tradeCount":2720,"turnover":38.18313115650708},"verdict":"rejected-no-edge","reasons":["consistency: beat the best baseline in only 0/8 windows (need 6)","multiplicity: mean excess Sharpe -0.413 < required 0.322 (bar after 3 tried candidates)"]}
+{"schema":"quant-arena.candidate.v1","at":"2026-07-22T21:02:50.071Z","candidateId":"cand-004-quant-researcher","proposer":"quant-researcher","authorModel":"sonnet","strategyPath":"/tmp/claude-1000/-home-drew-code-supervisor-lab/f06fd156-042a-4ef9-bd88-f2ec7f52b90c/scratchpad/qa-demo/strategies/cand-004-quant-researcher/strategy.ts","sha256":"sha256:8b7ecc3f848dbd5e6e101edca551cbdce2d7411fc5493ff30ce8c3ebc23be70b","authoringCostUsd":0.2651307,"nTried":4,"leakAudit":{"truncation":{"clean":true,"cutoffs":[121,514,906,1299,1691,2084],"divergence":null},"llm":{"verdict":"clean","evidence":""}},"eval":{"perWindow":[{"start":199,"end":703,"startDate":"2017-04-06","endDate":"2019-03-12","sharpe":0.9360771493648261,"bestBaselineSharpe":1.2094835481971722,"excess":-0.2734063988323461},{"start":468,"end":972,"startDate":"2018-04-18","endDate":"2020-03-23","sharpe":0.5593419153525067,"bestBaselineSharpe":0.6061551930118717,"excess":-0.04681327765936505},{"start":546,"end":1050,"startDate":"2018-08-06","endDate":"2020-07-09","sharpe":0.2027012985944738,"bestBaselineSharpe":0.2850058348026951,"excess":-0.08230453620822129},{"start":594,"end":1098,"startDate":"2018-10-11","endDate":"2020-09-15","sharpe":0.06825952465388957,"bestBaselineSharpe":0.17908115506948746,"excess":-0.1108216304155979},{"start":680,"end":1184,"startDate":"2019-02-08","endDate":"2021-01-13","sharpe":0.5277496931124905,"bestBaselineSharpe":0.660766081694157,"excess":-0.13301638858166642},{"start":1014,"end":1518,"startDate":"2020-05-21","endDate":"2022-04-26","sharpe":0.45263397984168435,"bestBaselineSharpe":0.5185595955384733,"excess":-0.06592561569678895},{"start":1057,"end":1561,"startDate":"2020-07-21","endDate":"2022-06-24","sharpe":0.4374222361779944,"bestBaselineSharpe":0.5223261328904791,"excess":-0.08490389671248472},{"start":1449,"end":1953,"startDate":"2022-01-20","endDate":"2023-12-26","sharpe":1.331981880357958,"bestBaselineSharpe":1.214848475647775,"excess":0.11713340471018285}],"meanExcessSharpe":-0.08500729242453595,"wins":1,"requiredWins":6,"threshold":0.3497663833473093},"inSampleFull":{"start":0,"end":2086,"days":2086,"totalReturn":1.8312941452636022,"maxDrawdown":0.19344761427023227,"sharpe":0.8783321625049441,"tradeCount":970,"turnover":10.48094104002003},"verdict":"rejected-no-edge","reasons":["consistency: beat the best baseline in only 1/8 windows (need 6)","multiplicity: mean excess Sharpe -0.085 < required 0.350 (bar after 4 tried candidates)"]}

package/src/quant-arena/fixtures/demo-campaign/rollout-manifest.json

@@ -0,0 +1,171 @@
+{
+  "schema": "quant-arena.rollout.v1",
+  "outDir": "/tmp/claude-1000/-home-drew-code-supervisor-lab/f06fd156-042a-4ef9-bd88-f2ec7f52b90c/scratchpad/qa-demo",
+  "at": "2026-07-22T21:05:12.016Z",
+  "notebookRows": 5,
+  "strategies": [
+    {
+      "label": "baseline-buy-hold-index",
+      "campaignDir": "/tmp/claude-1000/-home-drew-code-supervisor-lab/f06fd156-042a-4ef9-bd88-f2ec7f52b90c/scratchpad/qa-demo/campaign/baseline-buy-hold-index",
+      "cells": 8,
+      "scenarios": [
+        "window-1014-1518",
+        "window-1057-1561",
+        "window-1449-1953",
+        "window-199-703",
+        "window-468-972",
+        "window-546-1050",
+        "window-594-1098",
+        "window-680-1184"
+      ]
+    },
+    {
+      "label": "baseline-equal-weight",
+      "campaignDir": "/tmp/claude-1000/-home-drew-code-supervisor-lab/f06fd156-042a-4ef9-bd88-f2ec7f52b90c/scratchpad/qa-demo/campaign/baseline-equal-weight",
+      "cells": 8,
+      "scenarios": [
+        "window-1014-1518",
+        "window-1057-1561",
+        "window-1449-1953",
+        "window-199-703",
+        "window-468-972",
+        "window-546-1050",
+        "window-594-1098",
+        "window-680-1184"
+      ]
+    },
+    {
+      "label": "baseline-sma-crossover",
+      "campaignDir": "/tmp/claude-1000/-home-drew-code-supervisor-lab/f06fd156-042a-4ef9-bd88-f2ec7f52b90c/scratchpad/qa-demo/campaign/baseline-sma-crossover",
+      "cells": 8,
+      "scenarios": [
+        "window-1014-1518",
+        "window-1057-1561",
+        "window-1449-1953",
+        "window-199-703",
+        "window-468-972",
+        "window-546-1050",
+        "window-594-1098",
+        "window-680-1184"
+      ]
+    },
+    {
+      "label": "cand-001-default-author",
+      "campaignDir": "/tmp/claude-1000/-home-drew-code-supervisor-lab/f06fd156-042a-4ef9-bd88-f2ec7f52b90c/scratchpad/qa-demo/campaign/cand-001-default-author",
+      "cells": 8,
+      "scenarios": [
+        "window-1014-1518",
+        "window-1057-1561",
+        "window-1449-1953",
+        "window-199-703",
+        "window-468-972",
+        "window-546-1050",
+        "window-594-1098",
+        "window-680-1184"
+      ]
+    },
+    {
+      "label": "cand-002-default-author",
+      "campaignDir": "/tmp/claude-1000/-home-drew-code-supervisor-lab/f06fd156-042a-4ef9-bd88-f2ec7f52b90c/scratchpad/qa-demo/campaign/cand-002-default-author",
+      "cells": 8,
+      "scenarios": [
+        "window-1014-1518",
+        "window-1057-1561",
+        "window-1449-1953",
+        "window-199-703",
+        "window-468-972",
+        "window-546-1050",
+        "window-594-1098",
+        "window-680-1184"
+      ]
+    },
+    {
+      "label": "cand-003-quant-researcher",
+      "campaignDir": "/tmp/claude-1000/-home-drew-code-supervisor-lab/f06fd156-042a-4ef9-bd88-f2ec7f52b90c/scratchpad/qa-demo/campaign/cand-003-quant-researcher",
+      "cells": 8,
+      "scenarios": [
+        "window-1014-1518",
+        "window-1057-1561",
+        "window-1449-1953",
+        "window-199-703",
+        "window-468-972",
+        "window-546-1050",
+        "window-594-1098",
+        "window-680-1184"
+      ]
+    },
+    {
+      "label": "cand-004-quant-researcher",
+      "campaignDir": "/tmp/claude-1000/-home-drew-code-supervisor-lab/f06fd156-042a-4ef9-bd88-f2ec7f52b90c/scratchpad/qa-demo/campaign/cand-004-quant-researcher",
+      "cells": 8,
+      "scenarios": [
+        "window-1014-1518",
+        "window-1057-1561",
+        "window-1449-1953",
+        "window-199-703",
+        "window-468-972",
+        "window-546-1050",
+        "window-594-1098",
+        "window-680-1184"
+      ]
+    }
+  ],
+  "receipts": [
+    {
+      "callId": "f0af6aca-8e22-4264-91d0-81f1a391e345",
+      "phase": "search.proposal",
+      "actor": "proposer-shot:default-author",
+      "model": "sonnet",
+      "costUsd": 0.36238919999999997
+    },
+    {
+      "callId": "cc0c7780-83e5-4f6b-8cff-42ce8fcee086",
+      "phase": "audit.leak",
+      "actor": "leak-auditor:claude",
+      "model": "haiku",
+      "costUsd": 0.0626702
+    },
+    {
+      "callId": "15f76357-12e9-4164-a575-609ea19d6caf",
+      "phase": "search.proposal",
+      "actor": "proposer-shot:default-author",
+      "model": "sonnet",
+      "costUsd": 0.3580005
+    },
+    {
+      "callId": "544ff960-2698-4b41-a86e-4a43da853e15",
+      "phase": "audit.leak",
+      "actor": "leak-auditor:claude",
+      "model": "haiku",
+      "costUsd": 0.0432539
+    },
+    {
+      "callId": "7c240191-60d7-4322-868f-86dee8a8356b",
+      "phase": "search.proposal",
+      "actor": "proposer-shot:quant-researcher",
+      "model": "sonnet",
+      "costUsd": 0.4205679
+    },
+    {
+      "callId": "5492fe01-633d-4513-800b-194834546a77",
+      "phase": "audit.leak",
+      "actor": "leak-auditor:claude",
+      "model": "haiku",
+      "costUsd": 0.0799569
+    },
+    {
+      "callId": "d5f20214-c3f3-472a-ada6-130aeec8e76b",
+      "phase": "search.proposal",
+      "actor": "proposer-shot:quant-researcher",
+      "model": "sonnet",
+      "costUsd": 0.2651307
+    },
+    {
+      "callId": "a9f6a64b-2c17-4100-b1a1-cca2ce406ab7",
+      "phase": "audit.leak",
+      "actor": "leak-auditor:claude",
+      "model": "haiku",
+      "costUsd": 0.0571829
+    }
+  ]
+}

package/src/quant-arena/fixtures/demo-campaign/strategies/cand-001-default-author/strategy.ts

@@ -0,0 +1,119 @@
+type Bar = { date: string; open: number; high: number; low: number; close: number; volume: number };
+type Signal = { t: number; weights: number[] };
+
+export function generateSignals(bars: Bar[][]): Signal[] {
+  const N = bars.length;
+  const T = bars[0].length;
+
+  const SHORT_LB = 20;
+  const LONG_LB = 80;
+  const VOL_LB = 40;
+  const MIN_AVAIL = 25;
+  const REBAL_EVERY = 5;
+  const CHANGE_THRESHOLD = 0.05;
+  const MAX_WEIGHT = 0.30;
+  const TARGET_VOL = 0.15;
+  const MAX_GROSS = 0.95;
+  const MIN_SCALE = 0.15;
+  const TRADING_DAYS = 252;
+
+  const closesByTicker: number[][] = bars.map((series) => series.map((b) => b.close));
+
+  function smaAt(closes: number[], t: number, lb: number): number {
+    const start = Math.max(0, t - lb + 1);
+    let s = 0;
+    let n = 0;
+    for (let i = start; i <= t; i++) {
+      s += closes[i];
+      n++;
+    }
+    return n > 0 ? s / n : NaN;
+  }
+
+  function annualizedVol(closes: number[], t: number, lb: number): number {
+    const start = Math.max(1, t - lb + 1);
+    const rets: number[] = [];
+    for (let i = start; i <= t; i++) {
+      const prev = closes[i - 1];
+      const cur = closes[i];
+      if (prev > 0 && cur > 0) rets.push(Math.log(cur / prev));
+    }
+    if (rets.length < 2) return NaN;
+    let m = 0;
+    for (const r of rets) m += r;
+    m /= rets.length;
+    let v = 0;
+    for (const r of rets) v += (r - m) * (r - m);
+    v /= (rets.length - 1);
+    return Math.sqrt(v) * Math.sqrt(TRADING_DAYS);
+  }
+
+  const signals: Signal[] = [];
+  let lastWeights: number[] = new Array(N).fill(0);
+
+  for (let t = 0; t < T; t++) {
+    const avail = t + 1;
+    if (avail < MIN_AVAIL || t % REBAL_EVERY !== 0) continue;
+
+    const shortLb = Math.min(SHORT_LB, avail);
+    const longLb = Math.min(LONG_LB, avail);
+    const volLb = Math.min(VOL_LB, avail - 1);
+
+    const weights: number[] = new Array(N).fill(0);
+    const eligible: number[] = [];
+    const invVol: number[] = new Array(N).fill(0);
+
+    for (let k = 0; k < N; k++) {
+      const closes = closesByTicker[k];
+      const price = closes[t];
+      const smaShort = smaAt(closes, t, shortLb);
+      const smaLong = smaAt(closes, t, longLb);
+      const vol = annualizedVol(closes, t, volLb);
+      const trendUp = Number.isFinite(smaShort) && Number.isFinite(smaLong) && price > smaLong && smaShort > smaLong;
+      if (trendUp && Number.isFinite(vol) && vol > 1e-6) {
+        eligible.push(k);
+        invVol[k] = 1 / vol;
+      }
+    }
+
+    if (eligible.length > 0) {
+      const sumInv = eligible.reduce((s, k) => s + invVol[k], 0);
+      for (const k of eligible) weights[k] = invVol[k] / sumInv;
+
+      for (let iter = 0; iter < 5; iter++) {
+        let excess = 0;
+        const uncapped: number[] = [];
+        for (const k of eligible) {
+          if (weights[k] > MAX_WEIGHT) {
+            excess += weights[k] - MAX_WEIGHT;
+            weights[k] = MAX_WEIGHT;
+          } else {
+            uncapped.push(k);
+          }
+        }
+        if (excess <= 1e-9 || uncapped.length === 0) break;
+        const add = excess / uncapped.length;
+        for (const k of uncapped) weights[k] += add;
+      }
+
+      let basketVolSum = 0;
+      for (const k of eligible) basketVolSum += 1 / invVol[k];
+      const basketVol = basketVolSum / eligible.length;
+      let scale = basketVol > 0 ? TARGET_VOL / basketVol : MIN_SCALE;
+      if (!Number.isFinite(scale)) scale = MIN_SCALE;
+      scale = Math.max(MIN_SCALE, Math.min(1, scale));
+
+      for (const k of eligible) weights[k] *= scale * MAX_GROSS;
+    }
+
+    let diff = 0;
+    for (let k = 0; k < N; k++) diff += Math.abs(weights[k] - lastWeights[k]);
+
+    if (diff > CHANGE_THRESHOLD) {
+      signals.push({ t, weights: weights.slice() });
+      lastWeights = weights;
+    }
+  }
+
+  return signals;
+}

package/src/quant-arena/fixtures/demo-campaign/strategies/cand-002-default-author/strategy.ts

@@ -0,0 +1,119 @@
+type Bar = { date: string; open: number; high: number; low: number; close: number; volume: number };
+type Signal = { t: number; weights: number[] };
+
+export function generateSignals(bars: Bar[][]): Signal[] {
+  const nTickers = bars.length;
+  const nDays = bars[0].length;
+  const nStocks = nTickers - 1; // exclude index at k=0 (IDX) from direct allocation
+
+  const VOL_LOOKBACK = 20;
+  const MOM_LOOKBACK = 40;
+  const SMA_FAST = 20;
+  const SMA_SLOW = 60;
+  const MIN_HISTORY = SMA_SLOW;
+  const REBALANCE_INTERVAL = 21;
+  const VOL_TARGET = 0.12;
+  const TRADE_BAND = 0.04;
+  const URGENT_BAND = 0.15;
+
+  const signals: Signal[] = [];
+  if (nDays === 0 || nStocks <= 0) return signals;
+
+  const stdev = (rets: number[]): number => {
+    const n = rets.length;
+    if (n === 0) return 0;
+    let mean = 0;
+    for (let i = 0; i < n; i++) mean += rets[i];
+    mean /= n;
+    let v = 0;
+    for (let i = 0; i < n; i++) {
+      const d = rets[i] - mean;
+      v += d * d;
+    }
+    return Math.sqrt(v / n);
+  };
+
+  const sma = (k: number, t: number, period: number): number => {
+    let s = 0;
+    for (let i = t - period + 1; i <= t; i++) s += bars[k][i].close;
+    return s / period;
+  };
+
+  let lastWeights: number[] | null = null;
+  let lastTradeT = -Infinity;
+
+  for (let t = 0; t < nDays; t++) {
+    let target: number[] | null = null;
+
+    if (t === 0) {
+      target = new Array(nTickers).fill(0);
+      for (let k = 1; k <= nStocks; k++) target[k] = 1 / nStocks;
+    } else if (t >= MIN_HISTORY) {
+      const fast = sma(0, t, SMA_FAST);
+      const slow = sma(0, t, SMA_SLOW);
+      const uptrend = fast >= slow;
+      const regimeScale = uptrend ? 1.0 : 0.5;
+
+      const stats: { k: number; mom: number; vol: number }[] = [];
+      for (let k = 1; k <= nStocks; k++) {
+        const closes = bars[k];
+        const c0 = closes[t - MOM_LOOKBACK].close;
+        const c1 = closes[t].close;
+        const mom = c0 > 0 ? c1 / c0 - 1 : 0;
+
+        const rets: number[] = [];
+        for (let i = t - VOL_LOOKBACK + 1; i <= t; i++) {
+          const prev = closes[i - 1].close;
+          rets.push(prev > 0 ? closes[i].close / prev - 1 : 0);
+        }
+        const vol = stdev(rets);
+        stats.push({ k, mom, vol });
+      }
+
+      let pool = stats.filter((s) => s.mom > 0);
+      if (pool.length === 0) pool = stats;
+
+      const invVols = pool.map((s) => 1 / Math.max(s.vol, 1e-4));
+      const sumInv = invVols.reduce((a, b) => a + b, 0);
+      const rawWeights = pool.map((_, i) => invVols[i] / sumInv);
+
+      let weightedVol = 0;
+      for (let i = 0; i < pool.length; i++) weightedVol += rawWeights[i] * pool[i].vol;
+      const annVol = weightedVol * Math.sqrt(252);
+      let volScale = annVol > 0 ? VOL_TARGET / annVol : 1;
+      volScale = Math.min(volScale, 1);
+      volScale = Math.max(volScale, 0);
+
+      const combinedScale = Math.min(regimeScale, volScale);
+
+      target = new Array(nTickers).fill(0);
+      for (let i = 0; i < pool.length; i++) {
+        target[pool[i].k] = rawWeights[i] * combinedScale;
+      }
+    }
+
+    if (target === null) continue;
+
+    if (lastWeights === null) {
+      signals.push({ t, weights: target });
+      lastWeights = target;
+      lastTradeT = t;
+      continue;
+    }
+
+    let l1 = 0;
+    for (let k = 0; k < nTickers; k++) l1 += Math.abs(target[k] - lastWeights[k]);
+
+    const scheduled = t - lastTradeT >= REBALANCE_INTERVAL;
+    const urgent = l1 >= URGENT_BAND;
+    const worthTrading = l1 >= TRADE_BAND;
+
+    if ((scheduled && worthTrading) || urgent) {
+      signals.push({ t, weights: target });
+      lastWeights = target;
+      lastTradeT = t;
+    }
+  }
+
+  return signals;
+}

package/src/quant-arena/fixtures/demo-campaign/strategies/cand-003-quant-researcher/strategy.ts

@@ -0,0 +1,105 @@
+type Bar = { date: string; open: number; high: number; low: number; close: number; volume: number };
+type Signal = { t: number; weights: number[] };
+
+// ECONOMIC RATIONALE (one compound "regime quality" effect on a long-only equal-weight basket):
+//
+// 1) TREND BREADTH: capital is deployed in proportion to the fraction of the 10 names
+//    trading above their own trailing moving average. Broad participation in an uptrend
+//    is a well-documented leading indicator of continuation; breadth deteriorates BEFORE
+//    and DURING systemic drawdowns (fewer names hold their trend as a selloff spreads),
+//    so scaling exposure by breadth cuts risk ahead of/through the worst stretches without
+//    ever forecasting direction or shorting.
+//
+// 2) VOLATILITY TARGETING: equity returns cluster in volatility, and the conditional
+//    Sharpe ratio is empirically lower in high-vol regimes than in calm ones. Scaling
+//    exposure by (target vol / realized vol) keeps risk roughly constant through time,
+//    which mechanically improves the risk-adjusted return by underweighting exactly the
+//    stretches where the same dollar of exposure buys worse risk-adjusted payoff.
+//
+// Both signals are pure regime-quality throttles on ONE equal-weight book (no stock
+// picking, no shorting, no leverage - final scalar in [0,1]). They are combined
+// multiplicatively so the book only runs near-full size when trend AND vol conditions are
+// both benign, and de-risks fast (toward cash) when either deteriorates. A rebalance band
+// plus a minimum holding gap keep trading infrequent so 15bps one-way cost cannot eat the
+// edge - this strategy trades on regime shifts, not on daily noise.
+export function generateSignals(bars: Bar[][]): Signal[] {
+  const numTickers = bars.length;
+  const numStocks = numTickers - 1;
+  const numDays = bars[0].length;
+
+  const TREND_SMA = 100;      // trailing window defining "in an uptrend" per stock
+  const VOL_LOOKBACK = 20;    // trailing window for realized-vol estimate of the basket
+  const TARGET_VOL = 0.15;    // annualized vol target for the invested basket
+  const REBAL_BAND = 0.08;    // min change in target exposure to justify paying the spread
+  const MIN_GAP_DAYS = 5;     // minimum days between rebalances, avoids whipsaw churn
+
+  const signals: Signal[] = [];
+
+  const closeWindows: number[][] = [];
+  const closeSums: number[] = [];
+  for (let k = 0; k < numTickers; k++) {
+    closeWindows.push([]);
+    closeSums.push(0);
+  }
+
+  const basketReturns: number[] = [];
+
+  let lastScalar = -1;
+  let lastSignalDay = -Infinity;
+
+  for (let t = 0; t < numDays; t++) {
+    for (let k = 1; k < numTickers; k++) {
+      const c = bars[k][t].close;
+      closeWindows[k].push(c);
+      closeSums[k] += c;
+      if (closeWindows[k].length > TREND_SMA) {
+        closeSums[k] -= closeWindows[k].shift() as number;
+      }
+    }
+
+    if (t >= 1) {
+      let sumRet = 0;
+      for (let k = 1; k < numTickers; k++) {
+        const prev = bars[k][t - 1].close;
+        const cur = bars[k][t].close;
+        sumRet += prev > 0 ? (cur - prev) / prev : 0;
+      }
+      basketReturns.push(sumRet / numStocks);
+      if (basketReturns.length > VOL_LOOKBACK) basketReturns.shift();
+    }
+
+    const trendReady = closeWindows[1].length >= TREND_SMA;
+    const volReady = basketReturns.length >= VOL_LOOKBACK;
+    if (!trendReady || !volReady) continue;
+
+    let above = 0;
+    for (let k = 1; k < numTickers; k++) {
+      const sma = closeSums[k] / closeWindows[k].length;
+      if (bars[k][t].close > sma) above++;
+    }
+    const breadth = above / numStocks;
+
+    const mean = basketReturns.reduce((a, b) => a + b, 0) / basketReturns.length;
+    const variance =
+      basketReturns.reduce((a, b) => a + (b - mean) * (b - mean), 0) / basketReturns.length;
+    const annualizedVol = Math.sqrt(variance) * Math.sqrt(252);
+    const volScalar =
+      annualizedVol > 1e-8 ? Math.max(0, Math.min(1, TARGET_VOL / annualizedVol)) : 1;
+
+    const scalar = Math.max(0, Math.min(1, breadth * volScalar));
+
+    const daysSinceLast = t - lastSignalDay;
+    const moved = Math.abs(scalar - lastScalar) >= REBAL_BAND;
+
+    if (lastScalar < 0 || (moved && daysSinceLast >= MIN_GAP_DAYS)) {
+      const weights = new Array(numTickers).fill(0);
+      const perStock = scalar / numStocks;
+      for (let k = 1; k < numTickers; k++) weights[k] = perStock;
+      signals.push({ t, weights });
+      lastScalar = scalar;
+      lastSignalDay = t;
+    }
+  }
+
+  return signals;
+}