@tangle-network/agent-eval 0.163.2 → 0.171.0

package/docs/public-api.md

@@ -6,12 +6,12 @@ Generated by `pnpm api:census` on demand — this is a dated reading, not a gate
 
 | measure | count |
 | --- | --- |
-| export subpaths | 27 |
-| published value exports (subpath x symbol) | 1352 |
-| distinct symbols | 1183 |
-| production | 903 |
-| planned | 211 |
-| none | 238 |
+| export subpaths | 28 |
+| published value exports (subpath x symbol) | 1353 |
+| distinct symbols | 1184 |
+| production | 926 |
+| planned | 210 |
+| none | 217 |
 
 Type-only exports are not listed: a type binds no runtime surface, and removing one cannot break a caller at run time.
 
@@ -73,6 +73,13 @@ Every gap below makes a `none` less certain, so each one is answered by keeping
 
 A `none` row is a deletion candidate, not a deletion order. A symbol stays when removing it would lose something the census cannot weigh: a documented historical constant, a value another module of this package still needs, or a name whose only reference is a contract test over an on-disk artifact. Those are kept deliberately and stay listed here as `none`, so the next reader sees the same evidence and can decide again.
 
+The `none` set was reviewed symbol by symbol on 2026-08-21. Four rules decided most of it; each is recorded so the next reader inherits the judgment instead of redoing it.
+
+1. **A typed error constructor stays.** A caller discriminates a failure with `instanceof`, and a consumer that catches broadly binds no name for the import graph to see. Un-exporting one removes the only way to tell a named refusal from a bug, which contradicts this package's typed-outcome rule.
+2. **A wire schema or a contract-version constant stays.** `docs/wire-protocol.md`, `clients/python/`, and the OpenAPI document bind the same contract without binding a TypeScript symbol. That is blind spot 4 above, and it applies to every `*Schema`, `*_VERSION`, and `*_SCHEMA` row.
+3. **A published entry point with no doc is a documentation defect, not a dead export.** The per-harness coding-agent intake functions read `none` only because no Markdown front door named them; `docs/code-agent-intake.md` now does.
+4. **A symbol is deleted only when every channel is silent.** The 31-repository default-branch sweep, a GitHub code search across the organisation, the published `dist/` of every downstream package on npm, this repository's own source, tests, scripts, examples, and Markdown must all return nothing. Two symbols survived exactly this sweep: `runProposeReviewAsControlLoop`, whose caller is an uncommitted tool in `discovery-lab` but whose generated `.PROPOSAL.md` artifacts are committed there, and `shuffleOrder`, one of three stock contamination perturbations the changelog documents as a set.
+
 ## Exports
 
 ### `.`
@@ -246,7 +253,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `interRaterReliability` | production | agent-app:src/eval-campaign/trust-gate.ts |
 | `iqr` | production | this package: src/experiment-tracker.ts:27 |
 | `isBinaryOutcomeVector` | none | only this package's tests: src/statistics/paired-binary.test.ts:2 |
-| `isJudgeSpan` | production | this package: src/trace/query.ts:12 |
+| `isJudgeSpan` | production | this package: src/trace/query.ts:13 |
 | `isLlmSpan` | production | agent-runtime:src/candidate-execution/finalize.ts |
 | `isModelPriced` | production | agent-dev-container:products/intelligence/api/src/lib/optimization-engine.ts |
 | `isRunRecord` | production | this package: src/eval-trace-store.ts:21 |
@@ -400,13 +407,22 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `withJudgeRetry` | production | creative-agent:eval/lib/creative-judges.ts |
 | `wranglerDeployRunner` | production | blueprint-agent:scripts/experiments/lib/deploy-runner.ts |
 
+### `./adapters/http`
+
+2 value exports — 0 production, 2 planned, 0 none.
+
+| symbol | consumer | evidence |
+| --- | --- | --- |
+| `httpDispatch` | planned | example: examples/distributed-driver/driver.ts:20 |
+| `runDispatchServer` | planned | example: examples/distributed-driver/worker.ts:11 |
+
 ### `./analyst`
 
-131 value exports — 75 production, 16 planned, 40 none.
+127 value exports — 77 production, 19 planned, 31 none.
 
 | symbol | consumer | evidence |
 | --- | --- | --- |
-| `adaptPublicBenchmarkFindings` | production | this package: src/analyst/benchmark-public-rlm.ts:14 |
+| `adaptPublicBenchmarkFindings` | production | this package: scripts/gepa-analyst-campaign.ts:59 |
 | `AGENT_RX_UPSTREAM_REVISION` | production | this package: src/analyst/benchmark-command-result.ts:2 |
 | `agentRxBenchmarkCase` | production | this package: src/analyst/benchmark-public-data.ts:11 |
 | `agentRxPredictionsToFindings` | production | this package: src/analyst/benchmark-public-adapters.ts:3 |
@@ -414,10 +430,10 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM` | none | only this package's tests: src/analyst/benchmark-implementation.test.ts:7 |
 | `ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES` | none | only this package's tests: src/analyst/benchmark-implementation.test.ts:7 |
 | `ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256` | production | this package: src/analyst/benchmark-command-persistence.ts:27 |
-| `ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256` | none | only this package's tests: src/analyst/benchmark-reference-result.test.ts:6 |
-| `ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256` | none | only this package's tests: src/analyst/benchmark-reference-result.test.ts:6 |
+| `ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256` | planned | doc: .claude/skills/agent-eval/SKILL.md |
+| `ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256` | planned | doc: .claude/skills/agent-eval/SKILL.md |
 | `ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM` | none | only this package's tests: src/analyst/benchmark-implementation.test.ts:7 |
-| `ANALYST_BENCHMARK_IMPLEMENTATION_FILES` | none | only this package's tests: src/analyst/benchmark-implementation.test.ts:7 |
+| `ANALYST_BENCHMARK_IMPLEMENTATION_FILES` | planned | doc: .claude/skills/agent-eval/SKILL.md |
 | `ANALYST_BENCHMARK_IMPLEMENTATION_SHA256` | production | this package: src/analyst/benchmark-command-persistence.ts:27 |
 | `ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE` | production | this package: src/analyst/benchmark-command-persistence.ts:9 |
 | `ANALYST_BENCHMARK_MANIFEST_FILE` | production | this package: src/analyst/benchmark-command-persistence.ts:9 |
@@ -439,24 +455,21 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `buildPrimeRepairPrompt` | planned | doc: docs/prime-analyst.md |
 | `buildSkillUsageReport` | none | only this package's tests: src/analyst/kinds/skill-usage.test.ts:2 |
 | `buildTraceToolsForGroup` | production | this package: src/analyst/kind-factory.ts:15 |
-| `CODE_TRACE_BENCH_ANALYST_PROMPT` | production | this package: src/analyst/benchmark-runner-prime.ts:11 |
-| `codeTraceBenchCase` | production | this package: src/analyst/benchmark-public-data.ts:11 |
-| `codeTracerPredictionsToFindings` | none | only this package's tests: src/analyst/benchmark-datasets.test.ts:3 |
-| `coerceJson` | production | this package: src/analyst/finding-signature.ts:10 |
+| `CODE_TRACE_BENCH_ANALYST_PROMPT` | production | this package: scripts/gepa-analyst-campaign.ts:60 |
+| `codeTraceBenchCase` | production | this package: scripts/gepa-analyst-campaign.ts:58 |
+| `codeTracerPredictionsToFindings` | production | this package: benchmarks/trace-analysis/codetracebench-glm52-20260730/summarize-codetracer.mjs:6 |
+| `coerceJson` | production | this package: src/analyst/finding-codec.ts:19 |
 | `compareAnalystRunners` | production | this package: src/analyst/benchmark-command-result.ts:15 |
 | `computeFindingId` | production | agent-runtime:src/runtime/index.ts |
 | `CONTROL_INTEGRITY_ANALYST` | none | only this package's tests: src/analyst/kinds/control-integrity.test.ts:5 |
 | `ControlIntegrityAnalyst` | none | — |
 | `createChatClient` | production | agent-dev-container:products/intelligence/api/src/lib/intent-audit-analyst.ts |
 | `createDspyRlmTraceEngine` | production | agent-dev-container:products/intelligence/api/src/lib/deep-trace-analyst.ts |
-| `createJudgeAdapter` | none | — |
 | `createPrimeBenchmarkRunner` | production | this package: src/analyst/benchmark-command.ts:62 |
 | `createPublicBenchmarkDirectRunner` | production | this package: src/analyst/benchmark-command.ts:60 |
 | `createPublicBenchmarkRlmRunner` | production | this package: src/analyst/benchmark-command.ts:61 |
-| `createRunCriticAdapter` | none | — |
 | `createSemanticConceptJudgeAdapter` | none | only this package's tests: src/analyst/adapters.test.ts:2 |
 | `createTraceAnalyst` | production | blueprint-agent:scripts/experiments/lib/sandbox-driver/analyst-steer.ts |
-| `createVerifierAdapter` | none | — |
 | `decodeReplyRows` | production | this package: src/analyst/benchmark-public-model.ts:57 |
 | `DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES` | production | this package: src/analyst/benchmark-command.ts:73 |
 | `DEFAULT_TRACE_ANALYST_KINDS` | production | agent-dev-container:products/intelligence/api/src/lib/deep-trace-analyst.ts |
@@ -470,7 +483,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `emitSkillUsageFindings` | none | only this package's tests: src/analyst/kinds/skill-usage.test.ts:2 |
 | `emptyPrimeRawUsage` | none | only this package's tests: src/analyst/prime-protocol.test.ts:4 |
 | `emptyPublicBenchmarkRunner` | production | this package: src/analyst/benchmark-command.ts:62 |
-| `evidenceRefsFromRawFinding` | production | this package: src/analyst/benchmark-public-rlm.ts:36 |
+| `evidenceRefsFromRawFinding` | production | this package: scripts/gepa-analyst-campaign.ts:70 |
 | `ExactAnalystRunExecutionError` | planned | doc: docs/trace-analysis.md |
 | `expandCodeTraceFailureBlocks` | production | this package: src/analyst/benchmark-public-model.ts:29 |
 | `extractPrimeJsonObject` | none | only this package's tests: src/analyst/prime-protocol.test.ts:4 |
@@ -479,13 +492,13 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `FINDING_SUBJECT_SYNTAX` | none | only this package's tests: src/analyst/finding-subject.test.ts:2 |
 | `FindingsStore` | production | agent-builder:scripts/run-canonical-analyst-loop.ts |
 | `findingSubjectGrammarPromptFor` | production | this package: src/analyst/kinds/failure-mode.ts:38 |
-| `IMPROVEMENT_KIND_SPEC` | none | only this package's tests: src/analyst/finding-subject.test.ts:11 |
+| `IMPROVEMENT_KIND_SPEC` | none | only this package's tests: src/analyst/finding-subject.test.ts:13 |
 | `isProposalFinding` | none | only this package's tests: src/analyst/proposal-findings.test.ts:2 |
 | `KIND_EXPECTED_SUBJECTS` | production | this package: src/analyst/kind-factory.ts:14 |
-| `KNOWLEDGE_GAP_KIND_SPEC` | none | only this package's tests: src/analyst/finding-subject.test.ts:11 |
-| `KNOWLEDGE_POISONING_KIND_SPEC` | none | only this package's tests: src/analyst/finding-subject.test.ts:11 |
+| `KNOWLEDGE_GAP_KIND_SPEC` | none | only this package's tests: src/analyst/finding-subject.test.ts:13 |
+| `KNOWLEDGE_POISONING_KIND_SPEC` | none | only this package's tests: src/analyst/finding-subject.test.ts:13 |
 | `loadCodeTraceVerificationArtifacts` | production | this package: src/analyst/benchmark-public-data.ts:31 |
-| `loadPublicBenchmarkRows` | none | only this package's tests: src/analyst/benchmark-command-public-data.test.ts:13 |
+| `loadPublicBenchmarkRows` | production | this package: scripts/gepa-analyst-campaign.ts:64 |
 | `makeFinding` | production | agent-dev-container:products/intelligence/api/src/lib/consultant/candidates.ts |
 | `makeProposalFinding` | production | agent-runtime:bench/src/swe-arena/outer-loop.mts |
 | `MAX_INCORRECT_BLOCK_STEPS` | production | this package: src/analyst/benchmark-public-adapters.ts:9 |
@@ -498,7 +511,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `parseFindingSubject` | production | agent-dev-container:products/intelligence/api/src/lib/improvement-loop.ts |
 | `parseRawFinding` | production | this package: src/analyst/kind-factory.ts:7 |
 | `parseVerificationOutcome` | production | discovery-lab:tools/check-technical-campaign.mjs |
-| `preparePublicAnalystBenchmark` | production | this package: src/analyst/benchmark-command.ts:62 |
+| `preparePublicAnalystBenchmark` | production | this package: scripts/gepa-analyst-campaign.ts:64 |
 | `primeAnalystProtocolSha256` | planned | doc: docs/prime-analyst.md |
 | `primeCodeTraceAnalystDefinition` | planned | doc: docs/prime-analyst.md |
 | `primeProtocolSha256` | production | this package: src/analyst/benchmark-runner-prime.ts:19 |
@@ -506,13 +519,13 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `projectPrimeTrajectory` | production | this package: src/analyst/benchmark-runner-prime.ts:19 |
 | `publicBenchmarkDistributions` | none | — |
 | `publicBenchmarkProtocolSha256` | production | this package: src/analyst/benchmark-instructions-override.ts:2 |
-| `publicBenchmarkRlmInstructions` | production | this package: src/analyst/benchmark-public-rlm.ts:24 |
+| `publicBenchmarkRlmInstructions` | production | this package: scripts/gepa-analyst-campaign.ts:60 |
 | `publicBenchmarkSelectionReport` | none | only this package's tests: src/analyst/benchmark-command-public-data.test.ts:13 |
 | `publicBenchmarkSystemPrompt` | none | only this package's tests: src/analyst/benchmark-command-public-data.test.ts:13 |
 | `publicDirectAnalystDefinition` | planned | doc: docs/trace-analysis.md |
 | `publicRlmAnalystDefinition` | planned | doc: docs/trace-analysis.md |
 | `RAW_FINDING_SCHEMA_PROMPT` | production | this package: src/analyst/benchmark-public-rlm.ts:36 |
-| `RawAnalystFindingSchema` | production | this package: src/analyst/benchmark-public-rlm.ts:36 |
+| `RawAnalystFindingSchema` | production | this package: scripts/emit-finding-contract.ts:20 |
 | `readAnalystBenchmarkArtifact` | production | this package: src/analyst/benchmark-command.ts:42 |
 | `readAnalystInstructionsOverride` | production | this package: src/analyst/benchmark-command.ts:52 |
 | `registryBenchmarkRunner` | none | only this package's tests: src/analyst/benchmark.test.ts:3 |
@@ -524,17 +537,17 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `resolveTraceAnalystLimits` | production | this package: src/analyst/kind-factory.ts:5 |
 | `rlmEngineLimits` | none | only this package's tests: src/analyst/definition-parity.test.ts:26 |
 | `roundAgentRxStep` | production | this package: src/analyst/benchmark-agentrx-calibration.ts:2 |
-| `runAnalystBenchmark` | production | this package: src/analyst/benchmark-command.ts:12 |
+| `runAnalystBenchmark` | production | this package: benchmarks/trace-analysis/codetracebench-glm52-20260730/summarize-codetracer.mjs:6 |
 | `runAnalystBenchmarkCommand` | production | this package: src/cli.ts:16 |
 | `runPrimeExchange` | production | this package: src/analyst/benchmark-runner-prime.ts:19 |
-| `runTraceAnalyst` | production | this package: src/analyst/benchmark-public-rlm.ts:42 |
-| `scoreAnalystFindings` | production | this package: src/analyst/benchmark.ts:3 |
+| `runTraceAnalyst` | production | this package: scripts/gepa-analyst-campaign.ts:71 |
+| `scoreAnalystFindings` | production | this package: scripts/gepa-analyst-campaign.ts:57 |
 | `selectPublicBenchmarkRows` | none | only this package's tests: src/analyst/benchmark-command-public-data.test.ts:13 |
 | `SKILL_USAGE_ANALYST` | none | only this package's tests: src/analyst/kinds/skill-usage.test.ts:2 |
 | `stripCodeFences` | planned | named in agent-builder (bind not in the import graph) |
 | `summarizeAgentRxCalibration` | production | this package: src/analyst/benchmark-command-result.ts:2 |
 | `summarizeAnalystBenchmarkRunner` | production | this package: src/analyst/benchmark-command-result.ts:18 |
-| `summarizeCodeTraceCalibration` | production | this package: src/analyst/benchmark-command-result.ts:16 |
+| `summarizeCodeTraceCalibration` | production | this package: benchmarks/trace-analysis/codetracebench-glm52-20260730/summarize-codetracer.mjs:6 |
 | `traceStoreEvidenceResolver` | production | this package: src/analyst/benchmark-command.ts:12 |
 
 ### `./authenticity`
@@ -583,7 +596,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 
 ### `./campaign`
 
-117 value exports — 75 production, 15 planned, 27 none.
+117 value exports — 76 production, 15 planned, 26 none.
 
 | symbol | consumer | evidence |
 | --- | --- | --- |
@@ -696,7 +709,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `surfaceContentHash` | production | discovery-lab:tools/run-profile-confirmation.mjs |
 | `surfaceHash` | production | agent-builder:src/lib/.server/eval/loops/evolution-runner.ts |
 | `tangleTracesRoot` | none | only this package's tests: src/campaign/run-dir.test.ts:4 |
-| `traceAnalystQualityJudge` | none | only this package's tests: src/campaign/analyst-surface.test.ts:4 |
+| `traceAnalystQualityJudge` | production | this package: scripts/gepa-analyst-campaign.ts:74 |
 | `transientDispatchFailure` | planned | doc: docs/eval-surface-map.md |
 | `userStoryScoreboard` | production | blueprint-agent:scripts/experiments/eval/launch-scoreboard/run.ts |
 | `validateSearchLedgerEvent` | planned | consumer tests: discovery-lab:tools/search-ledger-planless.test.mjs |
@@ -707,7 +720,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 
 ### `./contract`
 
-64 value exports — 40 production, 6 planned, 18 none.
+64 value exports — 40 production, 15 planned, 9 none.
 
 | symbol | consumer | evidence |
 | --- | --- | --- |
@@ -728,13 +741,13 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `evaluatePairedMeasurements` | production | this package: src/contract/profile-measured-comparison.ts:26 |
 | `externalTextOptimizationMethod` | production | agent-app:src/eval-campaign/index.ts |
 | `FileSystemOutcomeStore` | production | phony:products/builder/api/src/eval/outcomes.ts |
-| `fromClaudeCodeSession` | none | only this package's tests: tests/contract-code-agent-intake.test.ts:6 |
-| `fromCodexSession` | none | only this package's tests: tests/contract-code-agent-intake.test.ts:6 |
+| `fromClaudeCodeSession` | planned | doc: docs/code-agent-intake.md |
+| `fromCodexSession` | planned | doc: docs/code-agent-intake.md |
 | `fromFeedbackTable` | planned | example: examples/customer-feedback-loop/index.ts:12 |
-| `fromKimiCodeSession` | none | only this package's tests: tests/contract-code-agent-intake.test.ts:6 |
-| `fromOpenCodeSession` | none | only this package's tests: tests/contract-code-agent-intake.test.ts:6 |
+| `fromKimiCodeSession` | planned | doc: docs/code-agent-intake.md |
+| `fromOpenCodeSession` | planned | doc: docs/code-agent-intake.md |
 | `fromOtelSpans` | production | agent-dev-container:packages/intelligence-kernel/src/deterministic.ts |
-| `fromPiSession` | none | only this package's tests: tests/contract-code-agent-intake.test.ts:6 |
+| `fromPiSession` | planned | doc: docs/code-agent-intake.md |
 | `fromRunRecordDir` | production | this package: src/contract/eval-reporting-suite.ts:25 |
 | `fsCampaignStorage` | production | agent-dev-container:products/intelligence/api/src/lib/eval-engine.ts |
 | `gepaOptimizationMethod` | production | agent-app:src/eval-campaign/index.ts |
@@ -748,10 +761,10 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `observeCodeAgentSession` | production | this package: src/contract/intake/code-agent-session.ts:13 |
 | `paretoPolicy` | none | only this package's tests: src/campaign/gates/promotion-policy.test.ts:3 |
 | `paretoSignificanceGate` | production | agent-app:src/eval-campaign/index.ts |
-| `parseAgentTrace` | none | only this package's tests: tests/intake-agent-trace.test.ts:3 |
-| `parseCodeAgentJsonl` | planned | named in discovery-lab (bind not in the import graph) |
-| `parseCodeAgentJsonlFile` | none | only this package's tests: tests/contract-code-agent-intake.test.ts:6 |
-| `partitionRunsByAuthoringModel` | none | only this package's tests: tests/intake-agent-trace.test.ts:3 |
+| `parseAgentTrace` | planned | doc: docs/code-agent-intake.md |
+| `parseCodeAgentJsonl` | planned | doc: docs/code-agent-intake.md |
+| `parseCodeAgentJsonlFile` | planned | doc: docs/code-agent-intake.md |
+| `partitionRunsByAuthoringModel` | planned | doc: docs/code-agent-intake.md |
 | `REFERENCE_EQUIVALENCE_INPUT_LIMITS` | none | only this package's tests: src/reference-equivalence-judge.test.ts:13 |
 | `REFERENCE_EQUIVALENCE_JUDGE_VERSION` | none | only this package's tests: src/reference-equivalence-judge.test.ts:13 |
 | `runAgentProfileImprovementExperiment` | production | agent-runtime:src/intelligence/authored-profile-improvement.ts |
@@ -769,7 +782,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `selfImprove` | production | agent-app:src/eval-campaign/index.ts |
 | `SelfImproveRunError` | production | agent-dev-container:products/intelligence/api/src/routes/optimizations.ts |
 | `skillOptOptimizationMethod` | production | agent-app:src/eval-campaign/index.ts |
-| `streamCodeAgentJsonlFile` | none | only this package's tests: tests/contract-code-agent-intake.test.ts:6 |
+| `streamCodeAgentJsonlFile` | planned | doc: docs/code-agent-intake.md |
 | `summarizeExecution` | production | traces:src/execution.ts |
 | `transientDispatchFailure` | planned | doc: docs/eval-surface-map.md |
 | `verifyAgentProfileImprovementExperimentComparison` | production | agent-runtime:src/intelligence/authored-profile-improvement.ts |
@@ -778,7 +791,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 
 ### `./experiment`
 
-83 value exports — 43 production, 22 planned, 18 none.
+83 value exports — 48 production, 18 planned, 17 none.
 
 | symbol | consumer | evidence |
 | --- | --- | --- |
@@ -792,13 +805,13 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `buildEvidenceVector` | planned | doc: docs/experiment.md |
 | `buildFunnel` | planned | doc: docs/experiment.md |
 | `classifyReissue` | none | only this package's tests: tests/experiment/preregistration-acceptance.test.ts:11 |
-| `clusteredPower` | planned | doc: docs/charter.md |
+| `clusteredPower` | production | this package: scripts/tb-gated-stop-ab.ts:55 |
 | `comparePairedArms` | production | agent-knowledge:src/memory/experiment/learning-metrics.ts |
 | `composeFunnels` | planned | doc: docs/experiment.md |
 | `computeEstimand` | production | this package: src/experiment/define.ts:21 |
 | `computeInterval` | production | this package: src/experiment/define.ts:21 |
 | `createEvidenceReceipt` | planned | consumer tests: agent-runtime:tests/integration/runtime-eval-pursuit-evidence.test.ts |
-| `defineExperiment` | planned | doc: docs/charter.md |
+| `defineExperiment` | production | this package: scripts/tb-gated-stop-ab.ts:55 |
 | `DesignRefusalError` | none | only this package's tests: tests/experiment/power.test.ts:10 |
 | `eProcess` | production | this package: src/campaign/gates/sequential.ts:35 |
 | `evaluateCondition` | planned | named in agent-dev-container (bind not in the import graph) |
@@ -831,7 +844,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `mcnemarPower` | production | supervisor-lab:bench/deepswe/headroom.ts |
 | `mcnemarRequiredN` | production | blueprint-agent:scripts/experiments/analyze-convergence.ts |
 | `mulberry32` | production | agent-knowledge:src/memory/holdout.ts |
-| `openSealedExperiment` | planned | example: examples/sealed-experiment/index.ts:12 |
+| `openSealedExperiment` | production | this package: scripts/tb-gated-stop-ab.ts:55 |
 | `pairArms` | production | agent-knowledge:src/memory/experiment/learning-pairs.ts |
 | `pairedBootstrap` | production | agent-builder:frontier/adapters/fleet-eval-runner.ts |
 | `pairedEvalueSequence` | production | creative-agent:src/lib/experiments/ab-design.ts |
@@ -847,13 +860,13 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `powerPreflight` | production | discovery-lab:tools/run-sequential-profile-improvement.mjs |
 | `projectNLadderBudget` | production | this package: src/experiment/define.ts:21 |
 | `readField` | planned | named in agent-dev-container (bind not in the import graph) |
-| `renderEvidenceIndex` | none | only this package's tests: src/experiment/evidence-record.test.ts:4 |
+| `renderEvidenceIndex` | production | this package: scripts/render-evidence-index.ts:15 |
 | `renderFunnelTable` | planned | example: examples/sealed-experiment/index.ts:12 |
 | `requiredPairedSampleSize` | production | discovery-lab:tools/design-gate.mjs |
 | `requiredSampleSize` | planned | doc: docs/design/statistics-decisions.md |
 | `runSelectionRule` | production | this package: src/experiment/define.ts:21 |
 | `runUniformPassBudget` | production | this package: src/experiment/define.ts:21 |
-| `sealExperiment` | planned | example: examples/sealed-experiment/index.ts:12 |
+| `sealExperiment` | production | this package: scripts/tb-gated-stop-ab.ts:55 |
 | `SealIntegrityError` | none | only this package's tests: tests/experiment/budget-and-seal.test.ts:8 |
 | `sequentialCrossingHorizon` | none | only this package's tests: tests/sequential.test.ts:2 |
 | `sequentialDecide` | planned | doc: docs/experiment.md |
@@ -872,13 +885,13 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 
 | symbol | consumer | evidence |
 | --- | --- | --- |
-| `adversarialObjective` | production | this package: src/fuzz/explorer.ts:22 |
+| `adversarialObjective` | production | this package: src/fuzz/explorer.ts:23 |
 | `BehaviorExplorer` | production | this package: src/fuzz/fuzz-agent.ts:9 |
-| `buildCapsule` | production | this package: src/fuzz/explorer.ts:19 |
+| `buildCapsule` | production | this package: src/fuzz/explorer.ts:20 |
 | `buildCoverage` | production | this package: src/fuzz/capsule.ts:12 |
 | `cellId` | planned | consumer tests: tax-agent:tests/eval/lib/fuzz-space.ts |
 | `composeGates` | production | creative-agent:eval/lib/fuzz-wiring.ts |
-| `enumerateCells` | production | this package: src/fuzz/explorer.ts:21 |
+| `enumerateCells` | production | this package: src/fuzz/explorer.ts:22 |
 | `fuzzAgent` | production | creative-agent:eval/fuzz.ts |
 | `makeExploreTools` | none | only this package's tests: src/fuzz/fuzz-agent.test.ts:8 |
 | `mutationProposer` | production | creative-agent:eval/fuzz.ts |
@@ -994,22 +1007,22 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `checkMultishotGolden` | planned | doc: docs/multishot-golden-records.md |
 | `checkMultishotGoldenScenario` | planned | doc: docs/multishot-golden-records.md |
 | `checkMultishotMatrixGoldenScenario` | planned | doc: docs/multishot-golden-records.md |
-| `compareJson` | production | this package: src/multishot/golden/harness.ts:10 |
+| `compareJson` | production | this package: scripts/record-multishot-golden.ts:32 |
 | `CURRENT_MULTISHOT_GOLDEN_VERSION` | planned | doc: docs/multishot-golden-records.md |
 | `goldenRecords` | production | this package: src/multishot/golden/harness.ts:23 |
 | `maskVolatileMarkdown` | none | only this package's tests: src/multishot/golden/golden.test.ts:20 |
 | `MultishotGoldenMismatchError` | planned | doc: docs/multishot-golden-records.md |
-| `multishotGoldenScenarios` | production | this package: src/multishot/golden/harness.ts:24 |
+| `multishotGoldenScenarios` | production | this package: scripts/record-multishot-golden.ts:49 |
 | `multishotGoldenVersions` | none | — |
-| `multishotMatrixGoldenScenarios` | production | this package: src/multishot/golden/harness.ts:12 |
-| `readRunDir` | production | this package: src/multishot/golden/harness.ts:16 |
-| `recordError` | production | this package: src/multishot/golden/harness.ts:16 |
+| `multishotMatrixGoldenScenarios` | production | this package: scripts/record-multishot-golden.ts:38 |
+| `readRunDir` | production | this package: scripts/record-multishot-golden.ts:42 |
+| `recordError` | production | this package: scripts/record-multishot-golden.ts:42 |
 | `recordJudgeRequest` | production | this package: src/multishot/golden/matrix-scenarios.ts:21 |
 | `recordMessage` | planned | named in agent-dev-container (bind not in the import graph) |
 | `recordRequest` | production | this package: src/multishot/golden/matrix-scenarios.ts:21 |
-| `recordResult` | production | this package: src/multishot/golden/harness.ts:16 |
-| `sortJudgeRequests` | production | this package: src/multishot/golden/harness.ts:16 |
-| `stripVolatile` | production | this package: src/multishot/golden/harness.ts:16 |
+| `recordResult` | production | this package: scripts/record-multishot-golden.ts:42 |
+| `sortJudgeRequests` | production | this package: scripts/record-multishot-golden.ts:42 |
+| `stripVolatile` | production | this package: scripts/record-multishot-golden.ts:42 |
 
 ### `./pipelines`
 
@@ -1302,7 +1315,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 
 ### `./trace-repair`
 
-96 value exports — 40 production, 31 planned, 25 none.
+96 value exports — 51 production, 23 planned, 22 none.
 
 | symbol | consumer | evidence |
 | --- | --- | --- |
@@ -1313,7 +1326,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `admissionArtifact` | planned | doc: docs/trace-repair-admission.md |
 | `AdmissionDenominatorError` | production | this package: src/trace-repair/admission.ts:36 |
 | `AdmissionIndependenceError` | none | only this package's tests: src/trace-repair/admission.test.ts:19 |
-| `admitRow` | production | this package: tests/trace-repair/fixtures.ts:11 |
+| `admitRow` | production | this package: scripts/tb-repair-milestone1.ts:41 |
 | `admittedCount` | planned | named in agent-knowledge (bind not in the import graph) |
 | `admittedRowIds` | none | only this package's tests: src/trace-repair/admission.test.ts:3 |
 | `askRepairArm` | planned | doc: docs/trace-repair-analyst-arms.md |
@@ -1328,7 +1341,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `blindTrajectory` | production | this package: src/trace-repair/analyst-arm.ts:41 |
 | `buildDenominatorChain` | production | this package: src/trace-repair/admission.ts:36 |
 | `checkInterventionBudget` | production | this package: src/trace-repair/analyst-arm.ts:33 |
-| `classifyActionPayload` | none | only this package's tests: tests/trace-repair/action-budget.test.ts:2 |
+| `classifyActionPayload` | production | this package: scripts/tb-repair-milestone1.ts:41 |
 | `continuationPolicyDigest` | production | this package: src/trace-repair/admission.ts:52 |
 | `ContinuationPolicyViolationError` | none | only this package's tests: src/trace-repair/continuation-policy.test.ts:11 |
 | `continuationSeed` | planned | doc: docs/trace-repair-continuation.md |
@@ -1337,39 +1350,39 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `controlCanRescue` | none | only this package's tests: tests/trace-repair/control-policy.test.ts:4 |
 | `countFunnel` | production | this package: src/trace-repair/delta-repair.ts:27 |
 | `createCompletionRepairArm` | planned | doc: docs/trace-repair-analyst-arms.md |
-| `createDockerContinuationEnvironment` | planned | doc: docs/trace-repair-continuation.md |
+| `createDockerContinuationEnvironment` | production | this package: scripts/tb-repair-freelunch.ts:32 |
 | `createDspyRepairArm` | planned | doc: docs/trace-repair-analyst-arms.md |
 | `CREDIT_TERMS` | none | only this package's tests: tests/trace-repair/degenerate-strategies.test.ts:8 |
-| `defineControlPolicy` | production | this package: tests/trace-repair/fixtures.ts:16 |
-| `definePinnedContinuationPolicy` | planned | doc: docs/trace-repair-admission.md |
+| `defineControlPolicy` | production | this package: scripts/tb-repair-milestone1.ts:41 |
+| `definePinnedContinuationPolicy` | production | this package: scripts/tb-repair-freelunch.ts:32 |
 | `DEGENERATE_STRATEGIES` | planned | doc: docs/trace-repair-grader.md |
-| `deltaRepair` | planned | doc: README.md |
+| `deltaRepair` | production | this package: scripts/tb-repair-milestone1.ts:41 |
 | `dockerRunArgs` | none | only this package's tests: src/trace-repair/docker-environment.test.ts:3 |
 | `DSPY_REPAIR_SIGNATURE` | planned | doc: docs/trace-repair-analyst-arms.md |
 | `DSPY_REPAIR_TASK_TOKEN` | none | only this package's tests: tests/trace-repair/arm-dspy.test.ts:19 |
 | `dspyRepairInstructions` | none | only this package's tests: tests/trace-repair/arm-dspy.test.ts:19 |
-| `gradeRepairRow` | planned | doc: docs/trace-repair-analyst-arms.md |
-| `injectedTestOracle` | planned | doc: docs/trace-repair-grader.md |
-| `isRecordedTimeout` | production | this package: src/trace-repair/grade.ts:51 |
-| `MINI_SWE_SYSTEM_MESSAGE` | production | this package: src/trace-repair/continuation-policy.ts:28 |
+| `gradeRepairRow` | production | this package: scripts/tb-repair-milestone1.ts:41 |
+| `injectedTestOracle` | production | this package: scripts/tb-gated-stop-ab.ts:62 |
+| `isRecordedTimeout` | production | this package: scripts/tb-repair-milestone1.ts:68 |
+| `MINI_SWE_SYSTEM_MESSAGE` | production | this package: scripts/tb-gated-stop-ab.ts:62 |
 | `NO_DECISIVE_FAILURE` | none | only this package's tests: tests/trace-repair/analyst-response.test.ts:2 |
 | `NO_OP_ACTIONS` | none | — |
-| `nodeProcessRunner` | planned | doc: docs/trace-repair-continuation.md |
+| `nodeProcessRunner` | production | this package: scripts/tb-repair-freelunch.ts:32 |
 | `NondeterministicOracleError` | none | only this package's tests: tests/trace-repair/oracle-determinism.test.ts:4 |
 | `noOpInjectionStep` | none | only this package's tests: src/trace-repair/admission.test.ts:3 |
 | `normalizeActionForComparison` | production | this package: src/trace-repair/grade.ts:31 |
-| `oracleDeterminism` | production | this package: tests/trace-repair/fixtures.ts:18 |
+| `oracleDeterminism` | production | this package: scripts/tb-oracle-determinism.ts:17 |
 | `OUTPUT_ELISION_THRESHOLD` | none | only this package's tests: src/trace-repair/mini-swe-scaffold.test.ts:2 |
-| `parseAction` | production | this package: src/trace-repair/continuation-policy.ts:28 |
+| `parseAction` | production | this package: scripts/tb-gated-stop-ab.ts:62 |
 | `parseAnalystResponse` | planned | doc: docs/trace-repair-grader.md |
-| `parseTaskOracleRegistry` | planned | doc: docs/trace-repair-admission.md |
+| `parseTaskOracleRegistry` | production | this package: scripts/tb-corpus-rows.ts:28 |
 | `renderAdmissionReport` | planned | doc: docs/trace-repair-admission.md |
 | `renderDeltaRepairReport` | planned | doc: docs/trace-repair-grader.md |
-| `renderFormatErrorObservation` | production | this package: src/trace-repair/continuation-policy.ts:28 |
-| `renderInstanceMessage` | production | this package: src/trace-repair/continuation-policy.ts:28 |
-| `renderObservation` | production | this package: src/trace-repair/continuation-policy.ts:28 |
+| `renderFormatErrorObservation` | production | this package: scripts/tb-gated-stop-ab.ts:62 |
+| `renderInstanceMessage` | production | this package: scripts/tb-gated-stop-ab.ts:62 |
+| `renderObservation` | production | this package: scripts/tb-gated-stop-ab.ts:62 |
 | `renderRepairTrajectory` | production | this package: src/trace-repair/arm-completion.ts:37 |
-| `renderTimeoutObservation` | production | this package: src/trace-repair/continuation-policy.ts:28 |
+| `renderTimeoutObservation` | production | this package: scripts/tb-gated-stop-ab.ts:62 |
 | `REPAIR_CONTRACT_LINES` | production | this package: src/trace-repair/arm-completion.ts:37 |
 | `REPAIR_QUESTION` | production | this package: src/trace-repair/arm-completion.ts:37 |
 | `REPAIR_REPAIR_CONTRACT_LINES` | production | this package: src/trace-repair/arm-completion.ts:37 |
@@ -1377,7 +1390,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `repairArmPromptSha256` | production | this package: src/trace-repair/analyst-arm.ts:42 |
 | `repairArmResponse` | planned | doc: docs/trace-repair-analyst-arms.md |
 | `repairCredit` | production | this package: src/trace-repair/grade.ts:40 |
-| `repairFinding` | none | only this package's tests: tests/trace-repair/analyst-response.test.ts:2 |
+| `repairFinding` | production | this package: scripts/tb-repair-milestone1.ts:41 |
 | `repairQuestionSha256` | production | this package: src/trace-repair/analyst-arm.ts:42 |
 | `repairTaskDefinition` | production | this package: src/trace-repair/arm-completion.ts:37 |
 | `repairTaskPolicy` | production | this package: src/trace-repair/arm-dspy.ts:48 |
@@ -1386,17 +1399,17 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `rolloutDigest` | planned | doc: docs/trace-repair-continuation.md |
 | `rolloutRecordedSteps` | planned | doc: docs/trace-repair-continuation.md |
 | `runAdmission` | planned | doc: docs/trace-repair-admission.md |
-| `runContinuation` | planned | doc: docs/trace-repair-continuation.md |
+| `runContinuation` | production | this package: scripts/tb-repair-freelunch.ts:32 |
 | `SCAFFOLD_INTERVENTION_BUDGET` | production | this package: src/trace-repair/arm-completion.ts:27 |
 | `scanShellAction` | none | only this package's tests: tests/trace-repair/action-budget.test.ts:2 |
-| `stratumOf` | production | this package: src/trace-repair/admission.ts:36 |
-| `submissionOf` | production | this package: src/trace-repair/continuation-policy.ts:28 |
+| `stratumOf` | production | this package: scripts/tb-gated-stop-ab.ts:62 |
+| `submissionOf` | production | this package: scripts/tb-gated-stop-ab.ts:62 |
 | `SUBMIT_SENTINEL` | production | this package: src/trace-repair/action-budget.ts:19 |
 | `SUITE_REWARD_UNIT` | none | only this package's tests: tests/trace-repair/oracle-determinism.test.ts:4 |
 | `taskOracleRegistry` | none | only this package's tests: src/trace-repair/admission-report.test.ts:7 |
-| `TB_REPAIR_ADMISSION_CRITERIA` | none | only this package's tests: tests/trace-repair/admission.test.ts:2 |
+| `TB_REPAIR_ADMISSION_CRITERIA` | production | this package: scripts/tb-repair-milestone1.ts:41 |
 | `TestOracleError` | none | only this package's tests: tests/trace-repair/test-oracle.test.ts:3 |
-| `testSuiteDigest` | production | this package: tests/trace-repair/fixtures.ts:34 |
+| `testSuiteDigest` | production | this package: scripts/tb-gated-stop-ab.ts:62 |
 | `TestSuiteTamperedError` | planned | doc: docs/trace-repair-grader.md |
 | `toRecordedSteps` | none | only this package's tests: src/trace-repair/continuation-policy.test.ts:25 |
 | `totalCost` | planned | doc: examples/adapt-a-text-optimizer/README.md |
@@ -1405,11 +1418,11 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 
 ### `./traces`
 
-112 value exports — 94 production, 7 planned, 11 none.
+115 value exports — 94 production, 7 planned, 14 none.
 
 | symbol | consumer | evidence |
 | --- | --- | --- |
-| `aggregateLlm` | production | this package: src/meta-eval/correlation-study.ts:14 |
+| `aggregateLlm` | none | only this package's tests: src/trace/query.test.ts:3 |
 | `analyzeTraces` | production | agent-builder:src/lib/.server/eval/analysts/canonical-trace-analyst.ts |
 | `applyLlmSpanOtlpAttributes` | production | traces:src/adapters/claude.ts |
 | `applyToolSpanOtlpAttributes` | production | braid:src/adapters/analysis/trace-store.ts |
@@ -1429,7 +1442,7 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `createOtelTracingStore` | none | only this package's tests: tests/trace-contracts.test.ts:4 |
 | `DEFAULT_REDACTION_RULES` | production | agent-dev-container:products/intelligence/api/src/lib/redact.ts |
 | `DEFAULT_TRACE_ANALYST_BUDGETS` | production | agent-builder:src/lib/.server/eval/stores/d1-trace-analysis-store-adapter.ts |
-| `defaultProviderRedactor` | production | this package: src/llm-client.ts:34 |
+| `defaultProviderRedactor` | production | this package: src/llm-client.ts:46 |
 | `defaultTraceInsightPanel` | none | only this package's tests: src/trace-analyst/insights.test.ts:3 |
 | `describeTraceInsightScope` | production | blueprint-agent:scripts/experiments/lib/analyze-vb-run/report-html.ts |
 | `domainEvidencePattern` | production | blueprint-agent:scripts/experiments/lib/analyze-vb-run/session-loader.ts |
@@ -1450,9 +1463,10 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `InMemoryRawProviderSink` | production | agent-builder:src/lib/.server/eval/loops/canonical-campaign.ts |
 | `InMemoryTraceStore` | production | agent-builder:src/lib/.server/eval/loops/auto-research-runner.ts |
 | `INPUT_VALUE` | production | braid:src/adapters/analysis/trace-event-projection.ts |
-| `isJudgeSpan` | production | this package: src/trace/query.ts:12 |
+| `isJudgeSpan` | production | this package: src/trace/query.ts:13 |
 | `isLlmSpan` | production | agent-runtime:src/candidate-execution/finalize.ts |
 | `isOtlpModelCall` | production | this package: src/contract/intake/otel-spans.ts:44 |
+| `isRunMetric` | none | — |
 | `isToolSpan` | production | agent-runtime:src/runtime/supervise/trace-evidence.ts |
 | `judgeSpans` | production | this package: src/builder-eval/three-layer-eval.ts:25 |
 | `LLM_CACHE_WRITE_TOKEN_ATTR_KEYS` | production | traces:src/file-export.ts |
@@ -1483,13 +1497,15 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 | `OUTPUT_VALUE` | production | agent-runtime:src/runtime/supervise-surface.ts |
 | `planTraceInsightQuestions` | none | only this package's tests: src/trace-analyst/insights.test.ts:3 |
 | `projectOtlpFlatLine` | production | this package: src/trace-analyst/otlp-to-run-records.ts:63 |
-| `providerFromBaseUrl` | production | this package: src/llm-client.ts:34 |
+| `providerFromBaseUrl` | production | this package: src/llm-client.ts:46 |
 | `REDACTION_VERSION` | production | agent-dev-container:products/intelligence/api/src/lib/redact.ts |
 | `redactString` | production | agent-dev-container:products/intelligence/api/src/lib/redact.ts |
 | `redactValue` | production | traces:src/index.ts |
 | `RUN_COST_ATTR_KEYS` | production | this package: src/trace/execution-measurements.ts:2 |
+| `RUN_METRICS` | none | — |
 | `runFailureClass` | production | starter-foundry:registry/layers/agent-eval/redteam/files/src/eval/redteam/runner.ts |
 | `RunIntegrityError` | production | creative-agent:eval/canonical-runner.ts |
+| `runMetricExtractor` | production | this package: src/meta-eval/calibration.ts:10 |
 | `runsForScenario` | production | phony:products/builder/api/src/eval/gate.ts |
 | `scoreTraceInsightReadiness` | production | blueprint-agent:scripts/experiments/lib/analyze-vb-run/report-html-sections.ts |
 | `SPAN_KIND_ATTR_KEYS` | production | traces:src/run-span-tree.ts |
@@ -1524,57 +1540,57 @@ A `none` row is a deletion candidate, not a deletion order. A symbol stays when
 
 ### `./trajectory-replay`
 
-49 value exports — 24 production, 19 planned, 6 none.
+49 value exports — 28 production, 16 planned, 5 none.
 
 | symbol | consumer | evidence |
 | --- | --- | --- |
 | `assertReplayableTrajectory` | none | only this package's tests: tests/trajectory-replay/steps.test.ts:7 |
 | `buildFixPrompt` | production | this package: src/trajectory-replay/fix-loop.ts:17 |
 | `buildRetryFixPrompt` | production | this package: src/trajectory-replay/fix-loop.ts:17 |
-| `classifyObservation` | planned | doc: docs/trajectory-replay.md |
+| `classifyObservation` | production | this package: scripts/tb-corpus-rows.ts:29 |
 | `classifyPrefixStep` | none | only this package's tests: tests/trajectory-replay/verify.test.ts:14 |
 | `classifyVerdict` | planned | named in agent-builder (bind not in the import graph) |
 | `clipText` | production | this package: src/trajectory-replay/fix-loop.ts:17 |
 | `countScriptCommands` | production | this package: src/trajectory-replay/fix-loop.ts:17 |
-| `decodeRecordedTurns` | planned | doc: docs/trajectory-replay.md |
+| `decodeRecordedTurns` | production | this package: scripts/tb-corpus-rows.ts:29 |
 | `derivedImageTag` | planned | named in traces (bind not in the import graph) |
 | `deriveFailureSignature` | production | this package: src/trace-repair/grade.ts:25 |
-| `dockerImagePreparer` | production | this package: src/trajectory-replay/batch.ts:41 |
-| `enumerateReplayableCases` | production | this package: src/trajectory-replay/batch.ts:32 |
+| `dockerImagePreparer` | production | this package: src/trajectory-replay/batch.ts:42 |
+| `enumerateReplayableCases` | production | this package: src/trajectory-replay/batch.ts:33 |
 | `extractFixCommand` | production | this package: src/trajectory-replay/fix-loop.ts:17 |
-| `finalRecordedOutcome` | planned | doc: docs/trajectory-replay.md |
+| `finalRecordedOutcome` | production | this package: scripts/tb-corpus-rows.ts:29 |
 | `findingReplayStep` | planned | named in traces (bind not in the import graph) |
 | `findingTrajectoryId` | planned | named in traces (bind not in the import graph) |
 | `FORMAT_ERROR_OBSERVATION_PREFIX` | production | this package: src/trace-repair/mini-swe-scaffold.ts:17 |
-| `generateFixCommand` | production | this package: src/trajectory-replay/batch.ts:39 |
+| `generateFixCommand` | production | this package: src/trajectory-replay/batch.ts:40 |
 | `goldIncorrectSteps` | planned | named in traces (bind not in the import graph) |
-| `ingestRecordedTrajectory` | production | this package: src/trajectory-replay/batch.ts:42 |
+| `ingestRecordedTrajectory` | production | this package: src/trajectory-replay/batch.ts:43 |
 | `isElidedField` | none | only this package's tests: tests/trajectory-replay/steps.test.ts:7 |
-| `isRecordedTimeout` | production | this package: src/trace-repair/grade.ts:51 |
+| `isRecordedTimeout` | production | this package: scripts/tb-repair-milestone1.ts:68 |
 | `isSubmitAction` | production | this package: src/trajectory-replay/corpus.ts:19 |
 | `isSubmitOnlyAction` | none | only this package's tests: tests/trajectory-replay/steps.test.ts:7 |
 | `parseCorpusFlag` | planned | named in traces (bind not in the import graph) |
 | `parseIncorrectStepsSubject` | production | this package: src/trajectory-replay/findings.ts:41 |
 | `parseObservationOutput` | production | this package: src/trajectory-replay/corpus.ts:19 |
-| `parseRecordedReturncode` | production | this package: src/trace-repair/grade.ts:25 |
-| `PREFIX_DIVERGENCE_TOLERANCE_PCT` | production | this package: src/trajectory-replay/batch.ts:42 |
+| `parseRecordedReturncode` | production | this package: scripts/tb-gated-stop-ab.ts:77 |
+| `PREFIX_DIVERGENCE_TOLERANCE_PCT` | production | this package: src/trajectory-replay/batch.ts:43 |
 | `readFindingsFile` | planned | named in traces (bind not in the import graph) |
 | `readLabelEntries` | planned | named in traces (bind not in the import graph) |
 | `renderBatchReport` | planned | named in traces (bind not in the import graph) |
 | `renderVerifiedFindingsSection` | planned | named in traces (bind not in the import graph) |
-| `replayVerify` | production | this package: src/trajectory-replay/batch.ts:42 |
+| `replayVerify` | production | this package: src/trajectory-replay/batch.ts:43 |
 | `replayVerifyFinding` | planned | doc: docs/trajectory-replay.md |
 | `resolveCaseResources` | production | this package: src/trajectory-replay/findings.ts:36 |
 | `resolveFindingInvocation` | planned | named in traces (bind not in the import graph) |
 | `resolveFindingReplayability` | planned | named in traces (bind not in the import graph) |
-| `runFixLoop` | production | this package: src/trajectory-replay/batch.ts:40 |
+| `runFixLoop` | production | this package: src/trajectory-replay/batch.ts:41 |
 | `runReplayBatch` | planned | doc: docs/trajectory-replay.md |
-| `SandboxCounterfactualRunner` | production | this package: src/trajectory-replay/batch.ts:42 |
+| `SandboxCounterfactualRunner` | production | this package: src/trajectory-replay/batch.ts:43 |
 | `seededSample` | planned | named in traces (bind not in the import graph) |
 | `SUBMIT_ACTION_SIGNATURE` | production | this package: src/trace-repair/mini-swe-scaffold.ts:17 |
 | `summarizePrefixReplay` | none | only this package's tests: tests/trajectory-replay/verify.test.ts:14 |
 | `TIMEOUT_OBSERVATION_MARKER` | production | this package: src/trace-repair/mini-swe-scaffold.ts:17 |
-| `unreadableExitCount` | none | only this package's tests: tests/trajectory-replay/steps.test.ts:7 |
+| `unreadableExitCount` | production | this package: scripts/tb-corpus-rows.ts:29 |
 | `verifyFindings` | planned | doc: docs/trajectory-replay.md |
 | `wrapActionForExec` | production | this package: src/trace-repair/grade.ts:24 |