Binding constraint: exit 1 in 521062ms: [buildTestCache] Pre-compiling and bundling 982 test suites with esbuild... ✔ active reserve marker validates pool and vault mints, then returns an executable RPC-vault mark (53.6322ms) ✔ active reserve marker fails closed when a vault account does n … e\test_bundle\markQuarantinePersistence.test.js 1.4mb .cache\test_bundle\e2ePipeline.test.js 1.4mb .cache\test_bundle\markIntegrity.test.js 1.4mb .cache\test_bundle\entryRuleGates.test.js 1.4mb ...and 977 more output files... Done in 3612ms FAILING (9): ✖ no NEW test file ships without a negative control | ✖ no NEW verdict-emitting checker reports clean over an unverified reference set | ✖ production census uses the real WaitOne-to-true-tail window and reports 30/1/4 = 35 | ✖ comments cannot impersonate the AST ownership anchors | ✖ removing the executable Start-Process -Wait lowers the direct count to 34 | ✖ bare and variable call operators inserted inside the hold each raise the count to 36 | ✖ moving the exact CI observer back inside the hold restores candidate 36 | ✖ a call operator after the true-tail release is outside the hold and leaves 35 unchanged | ✖ the independent heartbeat has no current-state write authority FIRST: AssertionError [ERR_ASSERTION]: A NEW test file asserts only happy-path outcomes. A suite that pins no FAILING case passes identically whether or not the property under test holds — it is one observation repeated, not tw CONTRACT: [test-contract] {"suite":"bot","files":982,"tests":10045,"durationMs":516615,"concurrency":4,"maxOldSpaceMb":1536,"perTestTimeoutMs":1800000,"wallTimeoutMs":3600000,"timedOut":false,"recursiveRunnerWarning":false,"childExit":1,"outcome":"TEST_FAILURE","orphanedChildren":0} | [test-contract] TEST_FAILURE: the suite ran to completion and at least one test failed — read the failure list above

{
  "profile": "FULL",
  "canonicalDataRoot": "C:\\Users\\Jeremy\\Cwod\\bankr-robinhood-agent\\data",
  "structuralFootprintIndex": {
    "name": "Structural Footprint Index v0",
    "authority": "DIAGNOSTIC_ONLY_NOT_READINESS"
  },
  "processMap": {
    "exitCode": 10,
    "durationMs": 18289,
    "outputDigest": "403637342f9505b4ef64a82185500fad87014c2ff2e70be99c9f8c9187f5ec3e"
  },
  "capabilityManifest": {
    "exitCode": 0,
    "durationMs": 1563,
    "outputDigest": "02960e1447fa7ddd661dace99faf0fd80950f11bd4e378e262ccd4c0aca9f98c"
  },
  "lint": {
    "exitCode": 0,
    "durationMs": 63077,
    "outputDigest": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
  },
  "gateDefinitionDigest": "0c30639c9cda913ede955d30376a8069c8d5e0a9a50fcd8a747cf8ec0af8908c",
  "externalEvidence": {
    "status": "VERIFIED",
    "sourcePath": "C:\\Users\\Jeremy\\Cwod\\bankr-robinhood-agent\\data\\receipts\\readiness-external-evidence.v1.jsonl",
    "physicalGeneration": "1180692972:9570149209213956:1788023854400.5332",
    "completeEof": 145016,
    "digest": "01e8b26951972056c5b55939be83f79d34790dcdd3c082e004f0d838da3998ec",
    "rowCount": 37,
    "detail": "strictly validated 37 content-addressed row(s)"
  }
}

1 · THE EXIT LEAK — 30 hard stops that should not have been reachable, -4.59 SOL. These trips reached a tape peak of at least 25% (median 77%), so the trail was ARMED and sat at 0.7x of peak — ABOVE the 0.5x-of-entry hard stop that nonetheless fired. For contrast the 387 hard stops that never armed cost -44.88 SOL — that is the rule working. This is the size of an execution gap, not a policy result, and the cause is undecided between engine downtime, peak-reset-on-restart, and rule ordering across a price gap. Folded live from 1070 trips in arm1-trade-autopsy.jsonl, 49m old · 16 hard-stop trips had no readable peak and are in neither bucket. Deeper: npx tsx src/cli/arm1ExitAnomalyCli.ts.

2 · THE RE-ENTRY LEAK — 61% of exits reclaimed +20% within 2h; cut-all-and-re-enter beats every partial. On the same 102 trips and the same post-exit paths: full-exit + re-entry +95.70 SOL (ex-top-1 +56.85) against the LIVE full-exit's +13.81 SOL (ex-top-1 −24.20); partial-75 +11.05, partial-50 +8.29, hold-all +2.77. UPPER BOUND — the re-entry arm buys at the next achievable print with no impact haircut and models no sell rule of its own, and 45 of 102 remainders never triggered the trail and are MARKED at the window's last price rather than realised. An unexited bag is a position, not a P&L. Producer src/cli/arm1ReentryStudyCli.ts — no artifact on disk, so this page cannot age these figures; they are stated as of the 2026-08-10 read and re-derive with npx tsx src/cli/arm1ReentryStudyCli.ts. An instrument with no artifact is invisible to every idle detector here, which is why it is listed above under on-demand readers.

Coverage is part of the result. Strict learning join: 3130 joined trips · 87 folded autopsy rows without oracle · 229 oracle rows without folded autopsy. This panel is the observed subset, not the whole ledger.

EXITS SKEW LATE — 290 too-late vs 66 too-early (4.4×), plus 288 bad in both directions. The book's exit failure is holding THROUGH the peak, not selling before it. Estimated replayable impact (best-policy benchmark, SOL): too-late cohort +36.85 SOL total (+0.13 SOL/trade) vs too-early cohort +0.27 SOL total (+~0.0041 SOL/trade).