@hecer/yoke 1.19.0 → 1.20.0

package/bench/results/codex-comparison-2026-09-29.json

@@ -0,0 +1,305 @@
+{
+  "auditedVersion": "1.19.0",
+  "auditedCommit": "05fac3db1a51412563005e07f80e6e5a9aab8e06",
+  "models": [
+    "gpt-6.1-sol"
+  ],
+  "effort": "low",
+  "groups": [
+    {
+      "fixture": "routing-queue",
+      "arm": "codex",
+      "runs": 2,
+      "accepted": 2,
+      "medianWallMs": 118278.5,
+      "medianFreshInputTokens": 25292,
+      "medianOutputTokens": 2162,
+      "medianAllInputTokens": 137548
+    },
+    {
+      "fixture": "routing-queue",
+      "arm": "yoke-serial",
+      "runs": 2,
+      "accepted": 2,
+      "medianWallMs": 274943.5,
+      "medianFreshInputTokens": 48524,
+      "medianOutputTokens": 4822,
+      "medianAllInputTokens": 396300
+    },
+    {
+      "fixture": "independent-utils",
+      "arm": "codex",
+      "runs": 1,
+      "accepted": 1,
+      "medianWallMs": 71283,
+      "medianFreshInputTokens": 10263,
+      "medianOutputTokens": 797,
+      "medianAllInputTokens": 84503
+    },
+    {
+      "fixture": "independent-utils",
+      "arm": "yoke-serial",
+      "runs": 1,
+      "accepted": 1,
+      "medianWallMs": 442189,
+      "medianFreshInputTokens": 92330,
+      "medianOutputTokens": 3905,
+      "medianAllInputTokens": 626986
+    },
+    {
+      "fixture": "independent-utils",
+      "arm": "yoke-parallel",
+      "runs": 1,
+      "accepted": 0,
+      "medianWallMs": 117101,
+      "medianFreshInputTokens": 81497,
+      "medianOutputTokens": 3487,
+      "medianAllInputTokens": 481753
+    }
+  ],
+  "runs": [
+    {
+      "fixture": "routing-queue",
+      "repeat": 1,
+      "arm": "codex",
+      "model": "gpt-6.1-sol",
+      "effort": "low",
+      "routing": false,
+      "nativeMultiAgent": false,
+      "wallMs": 112482,
+      "acceptanceMs": 175,
+      "accepted": true,
+      "outcome": "succeeded",
+      "usage": {
+        "inputTokens": 158620,
+        "cachedInputTokens": 121984,
+        "cacheWriteInputTokens": 0,
+        "outputTokens": 2190,
+        "reasoningOutputTokens": 18
+      },
+      "events": 0,
+      "freshInputTokens": 36636
+    },
+    {
+      "fixture": "routing-queue",
+      "repeat": 1,
+      "arm": "yoke-serial",
+      "model": "gpt-6.1-sol",
+      "effort": "low",
+      "routing": false,
+      "nativeMultiAgent": false,
+      "wallMs": 273000,
+      "acceptanceMs": 173,
+      "accepted": true,
+      "outcome": 0,
+      "usage": {
+        "inputTokens": 383543,
+        "cachedInputTokens": 334720,
+        "cacheWriteInputTokens": 0,
+        "outputTokens": 4924,
+        "reasoningOutputTokens": 104
+      },
+      "events": 11,
+      "freshInputTokens": 48823
+    },
+    {
+      "fixture": "routing-queue",
+      "repeat": 2,
+      "arm": "yoke-serial",
+      "model": "gpt-6.1-sol",
+      "effort": "low",
+      "routing": false,
+      "nativeMultiAgent": false,
+      "wallMs": 276887,
+      "acceptanceMs": 169,
+      "accepted": true,
+      "outcome": 0,
+      "usage": {
+        "inputTokens": 409057,
+        "cachedInputTokens": 360832,
+        "cacheWriteInputTokens": 0,
+        "outputTokens": 4720,
+        "reasoningOutputTokens": 81
+      },
+      "events": 11,
+      "freshInputTokens": 48225
+    },
+    {
+      "fixture": "routing-queue",
+      "repeat": 2,
+      "arm": "codex",
+      "model": "gpt-6.1-sol",
+      "effort": "low",
+      "routing": false,
+      "nativeMultiAgent": false,
+      "wallMs": 124075,
+      "acceptanceMs": 409,
+      "accepted": true,
+      "outcome": "succeeded",
+      "usage": {
+        "inputTokens": 116476,
+        "cachedInputTokens": 102528,
+        "cacheWriteInputTokens": 0,
+        "outputTokens": 2134,
+        "reasoningOutputTokens": 28
+      },
+      "events": 0,
+      "freshInputTokens": 13948
+    },
+    {
+      "fixture": "independent-utils",
+      "repeat": 1,
+      "arm": "codex",
+      "model": "gpt-6.1-sol",
+      "effort": "low",
+      "routing": false,
+      "nativeMultiAgent": false,
+      "parallel": 1,
+      "acceptanceDigest": "3e9734e6d286b7c9d19e6ce473eea17e1adf9e30a6ee8927c7240170602f14f7",
+      "wallMs": 71283,
+      "acceptanceMs": 354,
+      "accepted": true,
+      "outcome": "succeeded",
+      "usage": {
+        "inputTokens": 84503,
+        "cachedInputTokens": 74240,
+        "cacheWriteInputTokens": 0,
+        "outputTokens": 797,
+        "reasoningOutputTokens": 7
+      },
+      "events": 0,
+      "freshInputTokens": 10263
+    },
+    {
+      "fixture": "independent-utils",
+      "repeat": 1,
+      "arm": "yoke-serial",
+      "model": "gpt-6.1-sol",
+      "effort": "low",
+      "routing": false,
+      "nativeMultiAgent": false,
+      "parallel": 1,
+      "acceptanceDigest": "3e9734e6d286b7c9d19e6ce473eea17e1adf9e30a6ee8927c7240170602f14f7",
+      "wallMs": 442189,
+      "acceptanceMs": 454,
+      "accepted": true,
+      "outcome": 0,
+      "usage": {
+        "inputTokens": 626986,
+        "cachedInputTokens": 534656,
+        "cacheWriteInputTokens": 0,
+        "outputTokens": 3905,
+        "reasoningOutputTokens": 60
+      },
+      "events": 16,
+      "freshInputTokens": 92330
+    },
+    {
+      "fixture": "independent-utils",
+      "repeat": 1,
+      "arm": "yoke-parallel",
+      "model": "gpt-6.1-sol",
+      "effort": "low",
+      "routing": false,
+      "nativeMultiAgent": false,
+      "parallel": 3,
+      "acceptanceDigest": "3e9734e6d286b7c9d19e6ce473eea17e1adf9e30a6ee8927c7240170602f14f7",
+      "wallMs": 117101,
+      "acceptanceMs": 306,
+      "accepted": false,
+      "outcome": 1,
+      "usage": {
+        "inputTokens": 481753,
+        "cachedInputTokens": 400256,
+        "cacheWriteInputTokens": 0,
+        "outputTokens": 3487,
+        "reasoningOutputTokens": 72
+      },
+      "events": 22,
+      "freshInputTokens": 81497
+    }
+  ],
+  "limitations": [
+    "Small fixture sample; not representative of production projects",
+    "CPU and RAM not measured",
+    "Immutable tests replayed after execution; not hidden from agents",
+    "User config disabled; not an audit of every installed plugin",
+    "Routing disabled; fixed model and effort",
+    "Queue fixture: two alternating pairs; independent fixture: one sample per arm",
+    "Single Windows machine; uncontrolled background load",
+    "Failed parallel run is not an accepted speedup",
+    "No causal A/B test isolating context startup, worktree or verification overhead"
+  ],
+  "pricesMeasured": false,
+  "cpuMeasured": false,
+  "memoryMeasured": false,
+  "date": "2026-09-29",
+  "setupDiagnostics": [
+    {
+      "fixture": "independent-utils",
+      "repeat": 1,
+      "arm": "codex",
+      "model": "gpt-6.1-sol",
+      "effort": "low",
+      "routing": false,
+      "nativeMultiAgent": false,
+      "parallel": 1,
+      "wallMs": 75037,
+      "acceptanceMs": 352,
+      "accepted": true,
+      "outcome": "succeeded",
+      "usage": {
+        "inputTokens": 86851,
+        "cachedInputTokens": 75520,
+        "cacheWriteInputTokens": 0,
+        "outputTokens": 1013,
+        "reasoningOutputTokens": 37
+      },
+      "events": 0,
+      "diagnosticOnly": true,
+      "exclusionReason": "Initial fixture lacked runtime ignore rules; not grouped with corrected fixture"
+    },
+    {
+      "fixture": "independent-utils",
+      "repeat": 1,
+      "arm": "yoke-serial",
+      "model": "gpt-6.1-sol",
+      "effort": "low",
+      "routing": false,
+      "nativeMultiAgent": false,
+      "parallel": 1,
+      "wallMs": 116055,
+      "acceptanceMs": 627,
+      "accepted": false,
+      "outcome": 1,
+      "usage": {
+        "inputTokens": 176111,
+        "cachedInputTokens": 147072,
+        "cacheWriteInputTokens": 0,
+        "outputTokens": 1317,
+        "reasoningOutputTokens": 26
+      },
+      "events": 6,
+      "diagnosticOnly": true,
+      "exclusionReason": "Initial fixture lacked runtime ignore rules; not grouped with corrected fixture"
+    },
+    {
+      "fixture": "independent-utils",
+      "repeat": 1,
+      "arm": "yoke-parallel",
+      "model": "gpt-6.1-sol",
+      "effort": "low",
+      "routing": false,
+      "nativeMultiAgent": false,
+      "parallel": 3,
+      "wallMs": 3555,
+      "acceptanceMs": 678,
+      "accepted": false,
+      "outcome": 1,
+      "usage": null,
+      "events": 0,
+      "diagnosticOnly": true,
+      "exclusionReason": "Git worktree path exceeded Windows limit before provider startup"
+    }
+  ]
+}

package/bench/results/goal-integration-probes-2026-09-29.json

@@ -0,0 +1,107 @@
+{
+  "date": "2026-09-29",
+  "auditedVersion": "1.19.0",
+  "codexVersion": "0.161.0-alpha.2",
+  "poolProbe": {
+    "fakeExecutor": true,
+    "limit": 1,
+    "executedWhilePoolFull": true,
+    "goalStatus": "complete"
+  },
+  "budgetProbe": {
+    "fakeExecutor": true,
+    "tokenBudget": 1,
+    "measuredTokens": 100,
+    "timeBudgetMs": 60,
+    "totalWallMs": 652,
+    "recordedAgentMs": 3,
+    "status": "complete"
+  },
+  "objectiveProbe": {
+    "requestedArtifact": "new-feature.txt",
+    "requestedArtifactExists": false,
+    "acceptance": "only checks old-feature.txt",
+    "executorCalled": false,
+    "attempts": 0,
+    "status": "complete"
+  },
+  "nativeProtocolProbe": {
+    "modelDevelopmentTurnRequested": false,
+    "archived": true,
+    "calls": [
+      {
+        "method": "thread/goal/set",
+        "requestedStatus": null,
+        "returnedStatus": "active",
+        "tokenBudget": 1000,
+        "tokensUsed": 0
+      },
+      {
+        "method": "thread/goal/get",
+        "requestedStatus": null,
+        "returnedStatus": "active",
+        "tokenBudget": 1000,
+        "tokensUsed": 0
+      },
+      {
+        "method": "thread/goal/set",
+        "requestedStatus": "paused",
+        "returnedStatus": "paused",
+        "tokenBudget": 1000,
+        "tokensUsed": 0
+      },
+      {
+        "method": "thread/goal/get",
+        "requestedStatus": null,
+        "returnedStatus": "paused",
+        "tokenBudget": 1000,
+        "tokensUsed": 0
+      },
+      {
+        "method": "thread/goal/clear",
+        "requestedStatus": null,
+        "returnedStatus": null,
+        "tokenBudget": null,
+        "tokensUsed": null,
+        "cleared": true
+      },
+      {
+        "method": "thread/goal/get",
+        "requestedStatus": null,
+        "returnedStatus": null,
+        "tokenBudget": null,
+        "tokensUsed": null
+      }
+    ]
+  },
+  "parallelGateProbe": [
+    {
+      "fakeAgentsAndGit": true,
+      "requiresStoryContext": false,
+      "exitCode": 0,
+      "observedContexts": [
+        "A",
+        "B",
+        "C",
+        null,
+        null,
+        null
+      ],
+      "acceptedStories": 3
+    },
+    {
+      "fakeAgentsAndGit": true,
+      "requiresStoryContext": true,
+      "exitCode": 1,
+      "observedContexts": [
+        "A",
+        "B",
+        "C",
+        null,
+        null,
+        null
+      ],
+      "acceptedStories": 0
+    }
+  ]
+}

package/bench/results/native-goal-validation-2026-09-30.json

@@ -0,0 +1,122 @@
+{
+  "schemaVersion": 1,
+  "validationDate": "2026-09-30",
+  "generatedBy": "Codex-assisted validation",
+  "codexVersion": "0.161.0-alpha.2",
+  "model": "gpt-6.1-sol",
+  "reasoningEffort": "low",
+  "evidence": "Protected executable acceptance commands, saved provider counters, and shared capacity snapshots.",
+  "cases": {
+    "nativeBareStartup": {
+      "status": "blocked",
+      "limitation": "Codex app-server does not support ignoring user config. Native bare mode is now rejected before spawn."
+    },
+    "nativeBeforeWindowsShellFix": {
+      "status": "blocked",
+      "artifactCorrect": true,
+      "limitation": "Windows Store PowerShell repeatedly failed sandbox command launch; the two-minute agent deadline interrupted the turn and its usage remained unknown."
+    },
+    "retainedThreadResume": {
+      "status": "complete",
+      "sameGoalId": true,
+      "sameNativeThread": true,
+      "attemptsBefore": 1,
+      "attemptsAfter": 1,
+      "independentCheckStatus": "passed",
+      "explicitDisposableBudgetOverride": {
+        "clearTokenBudget": true,
+        "reason": "Historical interrupted turn had unknown usage; retained as unknown, previous report preserved"
+      }
+    },
+    "nativeAfterShellFixBeforeAccountingFix": {
+      "status": "complete",
+      "independentCheckStatus": "passed",
+      "reportedLastIncrement": {
+        "inputTokens": 23419,
+        "outputTokens": 60
+      },
+      "savedProviderCumulativeTotals": {
+        "inputTokens": 133408,
+        "outputTokens": 780,
+        "cachedInputTokens": 111232,
+        "reasoningOutputTokens": 56
+      },
+      "limitation": "This historical run initially counted only the last model request. The saved provider totals exposed the error; this report preserves both measurements."
+    },
+    "nativeAfterAccountingFix": {
+      "status": "complete",
+      "attempts": 1,
+      "attemptSuccess": true,
+      "nativeThreadBound": true,
+      "artifactCorrect": true,
+      "independentCheckStatus": "passed",
+      "agentDurationMs": 82634,
+      "wallDurationMs": 89855,
+      "limits": {
+        "maxAttempts": 1,
+        "agentMinutes": 5,
+        "wallMinutes": 7,
+        "tokenBudget": 300000
+      },
+      "measuredTurnUsage": {
+        "inputTokens": 153736,
+        "outputTokens": 606,
+        "cachedInputTokens": 134528,
+        "cacheWriteInputTokens": 0,
+        "reasoningOutputTokens": 53
+      },
+      "durableUsageBaselineMatchesTurnUsage": true,
+      "nativeGoalCounter": {
+        "tokensUsed": 0,
+        "interpretation": "The paused native goal counter remains separate from measured model usage."
+      },
+      "resume": {
+        "status": "complete",
+        "attempts": 1,
+        "sameGoalId": true,
+        "sameNativeThread": true
+      },
+      "cleanup": {
+        "remainingProviderRecords": 0,
+        "activeModelUnits": 0,
+        "activeCheckUnits": 0
+      }
+    },
+    "sharedCapacity": {
+      "fakeExecutor": true,
+      "limit": 1,
+      "queuedWhilePoolFull": true,
+      "executedWhilePoolFull": false,
+      "goalStatus": "complete"
+    },
+    "budgetOverrun": {
+      "fakeExecutor": true,
+      "tokenBudget": 1,
+      "measuredTokens": 100,
+      "timeBudgetMs": 60,
+      "totalWallMs": 1208,
+      "recordedAgentMs": 30,
+      "status": "blocked"
+    },
+    "unboundObjective": {
+      "requestedArtifact": "new-feature.txt",
+      "requestedArtifactExists": false,
+      "acceptance": "only checks old-feature.txt; intentionally unbound objective",
+      "executorCalled": false,
+      "attempts": 0,
+      "status": "blocked"
+    }
+  },
+  "accounting": {
+    "source": "thread/tokenUsage/updated.tokenUsage.total",
+    "method": "Subtract the durable same-thread cumulative baseline; new threads begin at zero. Repeated absolute snapshots are not added again.",
+    "unknownUsage": "Missing or malformed snapshots remain unknown. Decreasing counters mark the reused binding untrusted until a fresh thread is explicitly selected.",
+    "reference": "https://github.com/openai/symphony/blob/main/elixir/docs/token_accounting.md"
+  },
+  "limits": [
+    "These are bounded disposable smoke cases, not a long-running recovery or provider parity benchmark.",
+    "Native Codex goals remain experimental and opt-in; method-unavailable fallback does not hide authentication, network, or provider errors.",
+    "Native goals stay paused during explicit Yoke turns; independent executable checks determine Yoke completion.",
+    "Privacy exports omit thread identifiers, local root paths, and raw transcripts."
+  ]
+}

package/bench/results/parallel-validation-2026-09-30.json

@@ -0,0 +1,86 @@
+{
+  "schemaVersion": 1,
+  "date": "2026-09-30",
+  "purpose": "Single-arm parallel acceptance smoke validation; no baseline or efficiency comparison",
+  "aiDisclosure": "AI assisted report; authenticated AI providers implemented work inside disposable benchmark fixtures.",
+  "source": {
+    "version": "1.20.0",
+    "commit": "05fac3db1a51412563005e07f80e6e5a9aab8e06",
+    "state": "Uncommitted validation worktree; built before execution"
+  },
+  "fixture": "independent-utils",
+  "model": "gpt-6.1-sol",
+  "reasoningEffort": "low",
+  "arm": "yoke-parallel",
+  "parallelWorkers": 3,
+  "maxIterations": 3,
+  "routing": false,
+  "nativeMultiAgent": false,
+  "acceptanceDigest": "3e9734e6d286b7c9d19e6ce473eea17e1adf9e30a6ee8927c7240170602f14f7",
+  "samples": [
+    {
+      "sample": "initial-runtime-claims-failure",
+      "root": "<disposable-initial-smoke>",
+      "buildContext": "Before production runtime claim exclusions were added",
+      "version": "1.20.0",
+      "exitCode": 1,
+      "status": "blocked",
+      "accepted": false,
+      "durationMs": 148237,
+      "immutableAcceptanceDurationMs": 245,
+      "stories": { "total": 3, "accepted": 0, "rejectedIntegration": 3 },
+      "immutableTests": { "total": 15, "passed": 0, "failed": 15, "cancelled": 0, "skipped": 0 },
+      "usage": {
+        "inputTokens": 646600,
+        "cachedInputTokens": 571264,
+        "cacheWriteInputTokens": 0,
+        "outputTokens": 3796,
+        "reasoningOutputTokens": 91,
+        "costUsd": null
+      },
+      "measurement": { "measuredCalls": 3, "unknownCalls": 0, "unmeasuredAttempts": 3, "costAvailable": "unknown" },
+      "failure": {
+        "classification": "Runtime claim files missing from Git cleanliness exclusions",
+        "integrationReason": "target working tree is not clean",
+        "immutableReplayReason": "No implementations integrated; original source stubs remained unimplemented",
+        "diagnostic": "Built realGitOps.isClean returned true without a claim file and false with a temporary diagnostic claim file in the completed disposable fixture.",
+        "retainedCandidates": 3,
+        "modelCandidateQuality": "Not independently established by the final root replay"
+      }
+    },
+    {
+      "sample": "corrected-runtime-claims-smoke",
+      "root": "<disposable-corrected-smoke>",
+      "buildContext": "Production claim and worktree runtime exclusions added; rebuilt before execution",
+      "version": "1.20.0",
+      "exitCode": 0,
+      "status": "complete",
+      "accepted": true,
+      "durationMs": 112949,
+      "immutableAcceptanceDurationMs": 257,
+      "stories": { "total": 3, "accepted": 3, "rejectedIntegration": 0 },
+      "immutableTests": { "total": 15, "passed": 15, "failed": 0, "cancelled": 0, "skipped": 0, "nodeTestDurationMs": 159.3043 },
+      "immutableProtectedFiles": { "total": 4, "byteIdenticalToSeedAfterReplay": true },
+      "usage": {
+        "inputTokens": 456659,
+        "cachedInputTokens": 372096,
+        "cacheWriteInputTokens": 0,
+        "outputTokens": 3217,
+        "reasoningOutputTokens": 64,
+        "costUsd": null
+      },
+      "measurement": { "measuredCalls": 3, "unknownCalls": 0, "unmeasuredAttempts": 3, "costAvailable": "unknown" }
+    }
+  ],
+  "sampleLimits": [
+    "One initial failed parallel sample and one corrected parallel sample; each executed once",
+    "No baseline or serial arm; no general efficiency conclusion",
+    "Small fixture with three independent stories and fifteen visible tests",
+    "Immutable original verification files restored and replayed after provider execution",
+    "CPU and RAM not measured; machine load not controlled",
+    "Fixed model and low effort; routing and native multi-agent delegation disabled",
+    "User config disabled; installed plugins and host skill discovery not exhaustively audited",
+    "Reported token counters retained as observed; unknown cost is not zero cost",
+    "Private prompts, transcripts, identifiers and local root paths excluded; raw logs retained outside Git"
+  ]
+}

package/dist/agents/process-record-identity.js

@@ -0,0 +1,33 @@
+import { execFile } from 'node:child_process';
+import { writeFileSync, renameSync, existsSync, rmSync } from 'node:fs';
+import { randomUUID } from 'node:crypto';
+import { processIncarnation } from './process-incarnation.js';
+/** Resolve ownership without blocking cancellation timers on a slow CIM query. */
+export function trackProcessRecordIdentity(record, child) {
+    let cancelled = false;
+    const update = (identity) => {
+        if (!identity || cancelled || child.exitCode !== null || child.signalCode !== null || !record || !existsSync(record.path))
+            return;
+        const temporary = `${record.path}.${randomUUID()}.tmp`;
+        try {
+            writeFileSync(temporary, JSON.stringify({ version: record.version, owner: record.owner, targetDir: record.targetDir, childPid: record.childPid, startedAt: identity, workerId: record.workerId }), { flag: 'wx', mode: 0o600 });
+            renameSync(temporary, record.path);
+        }
+        catch { /* unknown identity remains conservative */ }
+        finally {
+            try {
+                rmSync(temporary, { force: true });
+            }
+            catch { /* stale temporary is diagnostic evidence */ }
+        }
+    };
+    if (!record || !child.pid)
+        return () => { };
+    if (process.platform !== 'win32') {
+        update(processIncarnation(child.pid));
+        return () => { cancelled = true; };
+    }
+    const query = execFile('powershell.exe', ['-NoProfile', '-NonInteractive', '-Command', '(Get-CimInstance Win32_Process -Filter "ProcessId = $env:YOKE_PROCESS_PID").CreationDate'], { env: { ...process.env, YOKE_PROCESS_PID: String(child.pid) }, windowsHide: true, timeout: 5000, maxBuffer: 4096 }, (error, stdout) => update(!error && stdout.trim() ? `win32:${stdout.trim()}` : undefined));
+    return () => { cancelled = true; if (query.exitCode === null && query.signalCode === null)
+        query.kill('SIGKILL'); };
+}