@tangle-network/agent-eval 0.140.1 → 0.141.0

package/CHANGELOG.md

@@ -4,6 +4,21 @@ All notable changes to `@tangle-network/agent-eval` and its sibling `agent-eval-
 
 ---
 
+## [0.141.0] - 2026-08-02 - portable trace identity and profile-native multishot
+
+### Changed
+
+- **Wire-format change: OTLP trace/span ids.** All three private paddings — `padTraceId` in `store-to-otlp` and `otel-export`, plus the same strip+pad body as `runToTraceId`/`padSpanId` in `otel.ts` — are retired for `deriveHexId` from `@tangle-network/agent-trace-contract` (new dependency), routed through one shared `src/trace/wire-ids.ts`. The exporters previously produced DIFFERENT trace ids for the same run, and the strip+pad family emitted invalid hex that embedded the raw run id in the wire id — flagged by the contract's own `non-hex-id` validator. Any id that is already a valid W3C id passes through unchanged (an inbound `traceparent` survives); every other id now derives to the contract id, so trace ids emitted by prior releases for the same run DO NOT match ids emitted by this one.
+- `MultishotShape.buildOpener` / `buildDriverSystemPrompt` are now OPTIONAL: omitted callbacks derive from the `AgentProfile` + persona payload (`defaultShapeFromProfile`, exported), so a pure-profile `runMultishot({ profile, persona })` works with no role-builder functions. Existing shapes keep working unchanged.
+
+### Deprecated
+
+- `AgentDriver`, `ProductClient`, `decideNextUserTurn`, `buildDriverSystemPrompt` — a one-product REST client and its persona-driver loop do not belong in the generic substrate; they move to the product repo / become a 2-node agent graph in the next major. Each warns on first use (once per process).
+
+### Removed
+
+- `buildWorkerDriverSystemPrompt` + `WorkerDriverContext` (zero callers in any package, re-measured). Its knowledge ships as seed DATA instead: `WORKER_DRIVER_DOCTRINE` (the "never write a thin steer" driving contract) and `HARNESS_BRIEFS` (per-harness capability + caveat briefs), so a driver profile can seed from it and an optimizer can improve it — a role expressed as a code function can never improve.
+
 ## [0.140.1] - 2026-07-31 - a supervisor journal is read, or reported unreadable
 
 ### Fixed

package/dist/analyst/index.d.ts

@@ -1,5 +1,5 @@
 import { a as MultiLayerVerifier, l as VerifyOptions, o as Severity } from "../multi-layer-verifier-BHY1gWAc.js";
-import { A as parseFindingSubject, C as FINDING_SUBJECT_KINDS, D as FindingSubjectStringSchema, E as FindingSubjectKind, F as DefineExactCustomAnalystOptions, G as SemanticConceptJudgeOptions, I as defineCustomAnalyst, L as defineTraceAnalyst, M as DspyRlmTraceEngineOptions, N as createDspyRlmTraceEngine, O as KIND_EXPECTED_SUBJECTS, P as DefineCustomAnalystOptions, S as FINDING_SUBJECT_GRAMMAR_PROMPT, T as FindingSubject, W as SemanticConceptJudgeInput, Y as RunCritic, Z as RunTrace, _ as FindingsDiff, a as SkillUsageScanConfig, b as defaultIsMaterial, c as DEFAULT_TRACE_ANALYST_KINDS, d as IMPROVEMENT_KIND_SPEC, f as FAILURE_MODE_KIND_SPEC, g as DiffPolicy, h as emitControlIntegrityFindings, i as SkillUsageReport, j as renderFindingSubject, k as findingSubjectGrammarPromptFor, l as KNOWLEDGE_POISONING_KIND_SPEC, m as ControlIntegrityAnalyst, n as SkillUsageAnalyst, o as buildSkillUsageReport, p as CONTROL_INTEGRITY_ANALYST, r as SkillUsageRecord, s as emitSkillUsageFindings, t as SKILL_USAGE_ANALYST, u as KNOWLEDGE_GAP_KIND_SPEC, v as FindingsStore, w as FINDING_SUBJECT_SYNTAX, x as diffFindings, y as PersistedFinding } from "../skill-usage-CVVnoIx-.js";
+import { A as parseFindingSubject, C as FINDING_SUBJECT_KINDS, D as FindingSubjectStringSchema, E as FindingSubjectKind, F as DefineExactCustomAnalystOptions, G as SemanticConceptJudgeOptions, I as defineCustomAnalyst, L as defineTraceAnalyst, M as DspyRlmTraceEngineOptions, N as createDspyRlmTraceEngine, O as KIND_EXPECTED_SUBJECTS, P as DefineCustomAnalystOptions, S as FINDING_SUBJECT_GRAMMAR_PROMPT, T as FindingSubject, W as SemanticConceptJudgeInput, Y as RunCritic, Z as RunTrace, _ as FindingsDiff, a as SkillUsageScanConfig, b as defaultIsMaterial, c as DEFAULT_TRACE_ANALYST_KINDS, d as IMPROVEMENT_KIND_SPEC, f as FAILURE_MODE_KIND_SPEC, g as DiffPolicy, h as emitControlIntegrityFindings, i as SkillUsageReport, j as renderFindingSubject, k as findingSubjectGrammarPromptFor, l as KNOWLEDGE_POISONING_KIND_SPEC, m as ControlIntegrityAnalyst, n as SkillUsageAnalyst, o as buildSkillUsageReport, p as CONTROL_INTEGRITY_ANALYST, r as SkillUsageRecord, s as emitSkillUsageFindings, t as SKILL_USAGE_ANALYST, u as KNOWLEDGE_GAP_KIND_SPEC, v as FindingsStore, w as FINDING_SUBJECT_SYNTAX, x as diffFindings, y as PersistedFinding } from "../skill-usage-CNZTk_LT.js";
 import { b as CustomTokenPricing, c as CostLedgerHandle } from "../cost-ledger-FuQvHxPm.js";
 import { A as ChatClient, B as SandboxSdkTransportOpts, F as CreateChatClientOpts, I as CustomTransportOpts, L as DirectProviderTransportOpts, M as ChatResponse, N as ChatTransport, P as CliBridgeTransportOpts, R as MockTransportOpts, V as createChatClient, j as ChatRequest, k as ChatCallOpts, m as JudgeInput, p as JudgeFn, z as RouterTransportOpts } from "../types-BjMFz88h.js";
 import { _ as makeFinding, a as AnalystInputKind, c as AnalystRunInputs, d as AnalystSeverity, f as AnalystUsageReceipt, g as computeFindingId, h as ProposalFindingOrigin, i as AnalystFinding, l as AnalystRunResult, m as ProposalFinding, n as AnalystContext, o as AnalystRequirements, p as EvidenceRef, r as AnalystCost, s as AnalystRunEvent, t as Analyst, u as AnalystRunSummary, v as makeProposalFinding, x as TraceAnalysisStore } from "../types-D3jh6F98.js";
@@ -269,6 +269,19 @@ declare function appendVerificationArtifactsToOtlp(otlpText: string, traceId: st
 //#endregion
 //#region src/analyst/benchmark-public-types.d.ts
 type PublicAnalystBenchmarkDataset = 'agentrx' | 'codetracebench';
+/**
+ * Caller-supplied replacement for the recursive runner's analyst instructions.
+ * Only the `dspy-rlm` runner accepts one; the direct runner rejects it, and the
+ * recursive runner's abstention fallback keeps the stock direct prompt. Every
+ * recorded protocol digest for an override run binds the stock protocol digest
+ * to `sha256`, so an override run is never confusable with a stock run.
+ */
+interface AnalystInstructionsOverride {
+  /** Complete instruction text used instead of the shipped RLM instructions. */
+  readonly text: string;
+  /** SHA-256 hex digest of `text`. */
+  readonly sha256: string;
+}
 interface PublicAnalystBenchmarkModelConfig {
   baseUrl: string;
   apiKey: string;
@@ -279,12 +292,20 @@ interface PublicAnalystBenchmarkModelConfig {
   pricing?: CustomTokenPricing;
   /** Independent per-case recursive-engine spend limit. Default: 1 USD. */
   maxCostUsdPerAnalysis?: number;
+  /** Replaces the shipped RLM instructions. `dspy-rlm` runner only. */
+  instructionsOverride?: AnalystInstructionsOverride;
   dspyRlm?: {
     runner?: ExternalOptimizerRunnerCommand;
     maxIterations?: number;
     maxLlmCalls?: number;
     maxToolCalls?: number;
     maxOutputChars?: number;
+    /**
+     * Independent engine runs per case. Above 1 (CodeTraceBench only), the
+     * runner scores the step-level majority consensus across all runs instead
+     * of a single draw. Default: 1.
+     */
+    samples?: number;
   };
   costLedger?: CostLedgerHandle;
   durability?: {
@@ -366,8 +387,22 @@ interface CodeTraceBlockDiagnostics {
   unresolvedBlockInteriorSteps: number[];
   /** Steps claimed by more than one block; the first block keeps the step. */
   overlappingBlockSteps: number[];
+  /** Blocks dropped for violating the protocol's width, order, or count limits. */
+  droppedBlocks: string[];
   /** Findings dropped before expansion because their shape or evidence is invalid. */
   rejectedFindings?: string[];
+  /** Out-of-block citations removed from findings that kept at least one in-block citation. */
+  trimmedCitations?: string[];
+}
+/**
+ * One expanded step with the accepted block that owns it. The expansion's
+ * per-step ownership record: exactly the steps that survived shape, count,
+ * and evidence checks, so consensus voting sees the same step set the
+ * benchmark scores.
+ */
+interface CodeTraceStepAssignment {
+  step: number;
+  block: CodeTraceFailureBlock;
 }
 declare function emptyPublicBenchmarkRunner(): AnalystBenchmarkRunner<AnalystRunInputs>;
 declare function adaptPublicBenchmarkFindings(options: {
@@ -380,6 +415,8 @@ declare function adaptPublicBenchmarkFindings(options: {
 }): Promise<{
   findings: AnalystFinding[];
   diagnostics: CodeTraceBlockDiagnostics | undefined;
+  /** Present for CodeTraceBench only; AgentRx has no step-level expansion. */
+  stepBlocks?: CodeTraceStepAssignment[];
 }>;
 /**
  * Expand contiguous failure blocks into one scored finding per member step.
@@ -398,6 +435,7 @@ declare function expandCodeTraceFailureBlocks(options: {
 }): Promise<{
   findings: AnalystFinding[];
   diagnostics: CodeTraceBlockDiagnostics;
+  stepBlocks: CodeTraceStepAssignment[];
 }>;
 //#endregion
 //#region src/analyst/benchmark-public-data.d.ts
@@ -428,7 +466,7 @@ declare const MAX_INCORRECT_BLOCK_STEPS = 12;
  * densest case carries 4 disjoint labeled blocks. Together with the per-block
  * cap this bounds one case at 192 predicted steps without a second ceiling. */
 declare const MAX_INCORRECT_BLOCKS = 16;
-declare const CODE_TRACE_BENCH_ANALYST_PROMPT = "Analyze exactly one coding-agent trajectory and its attached final verification.\nYour task is the CodeTraceBench incorrect-step task: identify every wrong state-changing action, bad hypothesis that drives an action, and regression.\nAn incorrect step remains incorrect when the agent later recovers or the final verification passes; a trajectory that ends in success still contains every mistake the agent made along the way.\nIncorrect steps occur in contiguous failure blocks: one mistake plus every consecutive following step that commits to, compounds, or acts on it.\nReport each failure block as exactly one finding whose first_step is the block's first incorrect step and whose last_step is its last, covering every consecutive step between them.\nSet first_step to the first step that commits the mistake, not the step that planned it and not a later step that repeats it.\nExtend last_step one step at a time, and only while the next step independently satisfies the incorrect-step definition on its own action and its own following observation.\nStop at the first step where the agent detects the problem, inspects it, or begins repairing it: a diagnostic probe, a test run that exposes the defect, or a repair action ends the block and is never inside it.\nA one-step block is a complete and correct answer.\nEvery step inside a block is scored on its own: naming a correct step costs exactly as much as missing an incorrect one, and naming only the first step of a longer block forfeits every unnamed step.\nReport blocks separated by at least one correct step as separate findings, and never let two blocks overlap.\nInspect the complete supplied trace data.\nUse the final-verification outcome as evidence about the final state, not as a rule for whether earlier steps were incorrect.\nFor each candidate block, inspect every assistant action in it and its following observation.\nAdmit a block only when you can point at the specific evidence it produced: name as consequence_step the step number whose action or observation shows the damage — a failing command, a wrong file state, a repeated failure, or rework the agent had to do because of this block. That step is the block's own last step when its observation already shows the damage, and a later step otherwise.\nWhen you cannot name that later step number from the trace you were given, drop the block; a plausible story about why a step looks wrong is not evidence that it was.\nJudge that consequence from the trajectory itself: a passing final verification is not evidence that a block caused nothing, and a failing final verification is not evidence that any particular block caused it.\nFor every block, decide whether the agent escaped the failure.\nMark escape_status \"escaped\" only when you can name the single later step that fully reversed the block, the agent needed no other step to recover, and nothing after that step revisits the same file, command, or hypothesis; write that step number in the rationale.\nMark escape_status \"unescaped\" in every other case, including whenever you are unsure.\nA passing final verification never makes a block escaped: the agent may have made the mistake and repaired it over several steps, and those steps are still incorrect.\nLabel a failed command when the assistant caused it through a wrong action or unsupported hypothesis.\nLabel the later corrective action only when that action is itself wrong.\nDo not label a diagnostic probe merely because it exposes an earlier defect.\nDo not label a redundant but correct read or search; CodeTraceBench scores unuseful steps separately, and this run scores incorrect steps only.\nDo not label a step solely because final verification failed.\nWhen final verification is unavailable, use only directly observed trajectory evidence.\nEvery step in a reported block MUST be the positive integer n from an existing assistant LLM span named step-<n>.\nNever select an EVALUATOR, TOOL, CHAIN, final-verification, benchmark-verification, or message-<n> span.\nBefore emitting a finding, inspect every covered span's attributes.content and describe only the actions shown there.\nReport at most 16 blocks and at most 12 steps in one block; when more candidates than that exist, report the ones you can support with the clearest downstream evidence.\nWhen the trajectory has no incorrect steps, return an empty findings array.";
+declare const CODE_TRACE_BENCH_ANALYST_PROMPT = "Analyze exactly one coding-agent trajectory and its attached final verification.\nYour task is the CodeTraceBench incorrect-step task: identify every incorrect step, defined as a wrong state-changing intervention given the evidence — a mislocalized edit, a wrong hypothesis that drives an action, a regression, an irrelevant change, or an incorrect dependency or configuration choice.\nIf the final verification failed, the trajectory MUST contain at least one incorrect step. Never return an empty findings array on a failing trajectory; trace backward until you find the root cause.\nWork backward, the way this benchmark was annotated, never by scanning forward for suspicious steps: start from the final verification outcome or the latest observed failure evidence, identify the immediately preceding step whose action or output produced that observed error, then recursively ask which earlier decision led to each intermediate failure, until the preceding steps contain no error or the cause is unrelated to the trajectory's own decisions.\nEach backward chain terminates at an error-critical step — the earliest decision that triggered the downstream cascade — and that step is the block's first_step: the step that committed the mistake, not the step that planned it and not a later step that repeats it.\nA block is a maximal contiguous sequence of strictly incorrect steps. A step belongs in the block ONLY if it introduces, propagates, or compounds the error. \nDo NOT include steps that merely \"act on\", diagnose, or react to the error. A diagnostic command, a test run exposing the bug, or a correct exploratory read is a CORRECT step. \nIf an incorrect step is followed by a correct diagnostic step and then another incorrect step, you MUST emit two separate blocks. NEVER bridge correct steps by grouping them into a single block with incorrect steps. Over-blocking drastically hurts your precision.\nAfter identifying first_step, extend last_step forward ONLY through consecutive steps that independently introduce, propagate, or compound the mistake. A cascade of repeated failed attempts at the same wrong approach is one maximal block, provided EVERY step is independently incorrect.\nDo not end a block merely because the agent tried a variation of the same wrong approach; a variation that still carries the error stays inside the block.\nA partially correct or ambiguous fix still counts as incorrect; the block ends only at the first step free of the error — a clean diagnostic read, the corrective action that closes the issue and needs no further rework, or a genuine abandonment of the wrong approach.\nBlock extent follows the traced chain and this forward extension, nothing else.\nReport each failure block as exactly one finding whose first_step is the block's first incorrect step and whose last_step is its last, covering every consecutive step between them.\nEvery step inside a block is scored on its own: naming a correct step costs exactly as much as missing an incorrect one, and naming only the first step of a longer block forfeits every unnamed step. Because of this, carefully verify every step between first_step and last_step. Only include steps that introduce, propagate, or compound the error.\nReport blocks separated by at least one correct step as separate findings, and never let two blocks overlap. If there are multiple separate failure cascades, emit a separate finding for each one.\nPrefer anchored blocks: a block whose chain traces back from observed failure evidence — a failing command or verification, an error observation, a regression, or, on a solved trajectory, a later step that reverts or supersedes it — outranks one without.\nWhen an action is clearly wrong on its own evidence but you cannot trace such an anchor, report the block anyway with proportionally lower confidence.\nA solved trajectory still carries every mistake made along the way: inspect its final patching and verification stages for a state-changing action that a later step reverted, superseded, or corrected — a wrong edit just before the final fix is incorrect even when every test ends green.\nBefore emitting a candidate block, check its boundaries.\nNeighbor check: ask whether the accusation fits one step earlier (the decision rather than its consequence) or one step later (the next step still acts on or reworks the same error) better than where you placed it, and move the boundary when it does; a boundary off by one step scores zero at that step.\nCompleteness check: a block must cover the maximal contiguous sequence of incorrect steps. If an agent fails at step 10, tries to fix it at 11, fails, and tries again at 12 and 13, all four steps are incorrect and must be included in the block. Never truncate a cascade. If you miss the later steps of a cascade, your recall drops to zero for them.\nCounterfactual check: ask which step's correct execution would have made the downstream failure or rework disappear, and move first_step onto that step; use this check only to move a boundary, never to delete a block.\nWidth check: inspect the step immediately before first_step and the step immediately after last_step; when either neighbor commits to, compounds, or propagates the same mistake, it belongs inside the block — move that boundary outward and repeat until both neighbors are free of the error.\nFor each block you keep, name as consequence_step the step number whose action or observation shows the damage — a failing command, a wrong file state, a repeated failure, or rework the agent had to do because of this block; that step is the block's own last step when its observation already shows the damage, and a later step otherwise.\nWhen you cannot name that step number from the trace you were given, drop the block; a plausible story about why a step looks wrong is not evidence that it was.\nA passing final verification is not evidence that a block caused nothing, and a failing final verification is not evidence that any particular block caused it.\nFor every block, decide whether the agent escaped the failure.\nMark escape_status \"escaped\" only when you can name the single later step that fully reversed the block, the agent needed no other step to recover, and nothing after that step revisits the same file, command, or hypothesis; write that step number in the rationale. If the agent required multiple steps to fix the issue, or if you are unsure, it is not escaped.\nMark escape_status \"unescaped\" in every other case, including whenever you are unsure.\nA passing final verification never makes a block escaped.\nDo not label a diagnostic probe or test run merely because its output exposes an earlier defect.\nDo not label a redundant but correct read or search; CodeTraceBench scores unuseful steps separately, and this run scores incorrect steps only. Never include a correct step in a block just to bridge two incorrect steps; instead, emit two separate blocks.\nDo not label a step solely because final verification failed.\nWhen final verification is unavailable, trace backward from the latest failure evidence inside the trajectory itself.\nEvery step in a reported block MUST be the positive integer n from an existing assistant LLM span named step-<n>.\nNever select an EVALUATOR, TOOL, CHAIN, final-verification, benchmark-verification, or message-<n> span.\nBefore emitting a finding, inspect every covered span's attributes.content and describe only the actions shown there.\nReport at most 16 blocks and at most 12 steps in one block; when more candidates than that exist, report the ones whose chains carry the clearest downstream evidence.\nWhen the trajectory has no incorrect steps — its final verification passed and the final-stage sweep found no reverted, superseded, or corrected action — return an empty findings array.";
 /** One-shot JSON transport prompt for the direct runner. */
 declare function publicBenchmarkSystemPrompt(dataset: PublicAnalystBenchmarkDataset): string;
 /** Tool-loop prompt for the recursive runner. Same task, subject-encoded block. */
@@ -545,12 +583,16 @@ interface AnalystBenchmarkArtifact {
     execution: {
       repetitions: number;
       concurrency: number;
+      /** Absent on artifacts produced before consensus sampling existed. */
+      rlmSamples?: number;
       model: string;
       maxOutputTokens: number;
       timeoutMs: number;
       maxCostUsd: number;
       maxArtifactBytes: number;
       analystProtocolSha256: string;
+      /** Present only when the run replaced the recursive analyst instructions. */
+      instructionsOverrideSha256?: string;
       implementationSha256: string;
       dependencyLockSha256: string;
     };
@@ -584,9 +626,13 @@ interface AnalystBenchmarkRunIdentity {
     seed: number;
     concurrency: number;
     repetitions: number;
+    /** Absent on manifests written before consensus sampling existed. */
+    rlmSamples?: number;
     maxCostUsd: number;
     maxArtifactBytes: number;
     analystProtocolSha256: string;
+    /** Present only when the run replaced the recursive analyst instructions. */
+    instructionsOverrideSha256?: string;
     implementationSha256: string;
     dependencyLockSha256: string;
     runnerIds: readonly ['empty', string];
@@ -680,6 +726,8 @@ interface AnalystBenchmarkCommandConfig {
   seed: number;
   concurrency: number;
   repetitions: number;
+  /** Recursive-engine runs per case; above 1 the consensus is scored. */
+  rlmSamples: number;
   maxCostUsd: number;
   maxArtifactBytes: number;
   apiKeyEnv: string;
@@ -687,13 +735,13 @@ interface AnalystBenchmarkCommandConfig {
   resume: boolean;
 }
 declare function runAnalystBenchmarkCommand(argv: readonly string[], env?: NodeJS.ProcessEnv, dependencies?: AnalystBenchmarkCommandDependencies): Promise<number>;
-declare const ANALYST_BENCHMARK_HELP = "agent-eval analyst-benchmark\n\nRun the recursive DSPy trace analyst against public AgentRx or CodeTraceBench labels.\n\nRequired:\n  --dataset agentrx|codetracebench\n  --analyst dspy-rlm|direct        Scored analyst. Default: dspy-rlm.\n                                   'direct' is the retired one-shot runner that\n                                   produced the published evidence.\n  --labels <dataset.json|dataset.jsonl>\n  --trace-dir <one-trace-per-file OTLP JSONL directory>\n  --artifact-dir <extracted artifact root>  Required for CodeTraceBench\n  --out <new output directory>\n  --revision <full 40- or 64-character hex digest>\n  --split <dataset split>\n  --base-url <OpenAI-compatible /v1 URL>\n  --api-key-env <environment variable containing the bearer>\n  --model <provider model id>\n  --limit <positive case count>\n\nControls:\n  --resume                         Continue an interrupted run in --out\n  --seed <integer>                 Case-selection and comparison seed. Default: 0\n  --concurrency <positive integer> Parallel benchmark jobs. Default: 1\n  --repetitions <positive integer> Runs per case and runner. Default: 1\n  --max-output-tokens <positive>   Model output limit per call. Default: 16384\n  --python <executable>             Python with agent-eval-rpc[dspy]. Default: python\n  --timeout-ms <positive>          Model analyst deadline per case. Default: 300000\n  --max-cost-usd <positive>        Run-wide spend limit. Default: 5\n  --max-artifact-bytes <positive>  Final evidence bytes per case. Default: 8388608\n\nWrites result.json with every observation, metric, usage field, error, comparison,\ninput digest, artifact digest, case distribution, selected case id, and explicit\nunknown cost. Limited deterministic-hash subsets are marked non-representative.\nCompleted observations are fsynced to observations.jsonl. Shareable output is in\nresult.json and report.md. Machine-local paths, endpoint, and command are isolated\nin run.local.json.\nThe key is read from the named environment variable and is never written.";
+declare const ANALYST_BENCHMARK_HELP = "agent-eval analyst-benchmark\n\nRun the recursive DSPy trace analyst against public AgentRx or CodeTraceBench labels.\n\nRequired:\n  --dataset agentrx|codetracebench\n  --analyst dspy-rlm|direct        Scored analyst. Default: dspy-rlm.\n                                   'direct' is the retired one-shot runner that\n                                   produced the published evidence.\n  --labels <dataset.json|dataset.jsonl>\n  --trace-dir <one-trace-per-file OTLP JSONL directory>\n  --artifact-dir <extracted artifact root>  Required for CodeTraceBench\n  --out <new output directory>\n  --revision <full 40- or 64-character hex digest>\n  --split <dataset split>\n  --base-url <OpenAI-compatible /v1 URL>\n  --api-key-env <environment variable containing the bearer>\n  --model <provider model id>\n  --limit <positive case count>\n\nControls:\n  --resume                         Continue an interrupted run in --out\n  --seed <integer>                 Case-selection and comparison seed. Default: 0\n  --concurrency <positive integer> Parallel benchmark jobs. Default: 1\n  --repetitions <positive integer> Runs per case and runner. Default: 1\n  --rlm-samples <positive integer> Recursive-engine runs per case; above 1 the\n                                   step-level majority consensus is scored\n                                   (CodeTraceBench + dspy-rlm only). Default: 1\n  --instructions-file <path>       Replace the recursive analyst instructions\n                                   with this file's text (dspy-rlm only). The\n                                   recorded protocol digest binds the stock\n                                   protocol to the override text, and\n                                   result.json records instructionsOverrideSha256.\n  --max-output-tokens <positive>   Model output limit per call. Default: 16384\n  --python <executable>             Python with agent-eval-rpc[dspy]. Default: python\n  --timeout-ms <positive>          Model analyst deadline per case. Default: 300000\n  --max-cost-usd <positive>        Run-wide spend limit. Default: 5\n  --max-artifact-bytes <positive>  Final evidence bytes per case. Default: 8388608\n\nWrites result.json with every observation, metric, usage field, error, comparison,\ninput digest, artifact digest, case distribution, selected case id, and explicit\nunknown cost. Limited deterministic-hash subsets are marked non-representative.\nCompleted observations are fsynced to observations.jsonl. Shareable output is in\nresult.json and report.md. Machine-local paths, endpoint, and command are isolated\nin run.local.json.\nThe key is read from the named environment variable and is never written.";
 //#endregion
 //#region src/analyst/benchmark-implementation.d.ts
 declare const ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM = "sha256-canonical-source-manifest";
 declare const ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM = "sha256-canonical-file-manifest";
 declare const ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES: readonly string[];
-declare const ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 = "701ea59184f3d227cf590ea3f16d24d65dbed2ad5644f8294801ceb66a30d4ef";
+declare const ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 = "106219b86a612a0352fabf44eb34bc1bbffd5c842ed97ae6132343af88f89c94";
 /** The published benchmark evidence was produced at this package version, by
  * the retired one-shot direct runner, before trace analysts moved to the
  * recursive DSPy RLM engine. Both evidence digests below are historical facts
@@ -705,13 +753,33 @@ declare const ANALYST_BENCHMARK_EVIDENCE_PACKAGE_VERSION = "0.137.0";
 declare const ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256 = "1e03f2daed356d60316aabefb407ec1e437ac94d408d61eea4ae096e9c6fbb5b";
 declare const ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256 = "4dba263b6256a30d56c7fdb2d992d3a953c0035d731f359b704db806f68f75ac";
 declare const ANALYST_BENCHMARK_IMPLEMENTATION_FILES: readonly string[];
-declare const ANALYST_BENCHMARK_IMPLEMENTATION_SHA256 = "56dd4c7ed19fc5855f99ad238464ad95cb1bea155ad58c49dab3eb0f6cbe7d6a";
+declare const ANALYST_BENCHMARK_IMPLEMENTATION_SHA256 = "a3f5a8203f0a7c74736170c53100c50c4db1cf84381533002f0539ef1e7a4c6d";
 declare function analystBenchmarkImplementationDigest(): string;
 declare function analystBenchmarkDependencyLockDigest(): string;
 //#endregion
+//#region src/analyst/benchmark-instructions-override.d.ts
+/** Build an override from instruction text. Blank text is a caller error. */
+declare function analystInstructionsOverrideFromText(text: string): AnalystInstructionsOverride;
+/** Read override instructions from a file. Any read failure is fatal. */
+declare function readAnalystInstructionsOverride(path: string): AnalystInstructionsOverride;
+/**
+ * Protocol digest of the run as executed.
+ *
+ * Without an override this is exactly `publicBenchmarkProtocolSha256(dataset)`,
+ * so stock runs stay byte-identical to runs recorded before the override
+ * existed. With an override the digest binds the stock protocol digest (which
+ * covers both shipped prompts, including the abstention fallback's direct
+ * prompt) to the exact override text, so the recorded digest always hashes the
+ * instructions that actually ran.
+ */
+declare function effectiveAnalystProtocolSha256(dataset: PublicAnalystBenchmarkDataset, override?: Pick<AnalystInstructionsOverride, 'sha256'>): string;
+//#endregion
 //#region src/analyst/benchmark-report.d.ts
 declare function renderAnalystBenchmarkMarkdown(result: AnalystBenchmarkResult, comparisons?: readonly AnalystRunnerComparison[]): string;
 //#endregion
+//#region src/analyst/benchmark-summary.d.ts
+declare function summarizeAnalystBenchmarkRunner(runnerId: string, observations: readonly AnalystBenchmarkObservation[]): AnalystBenchmarkSummary;
+//#endregion
 //#region src/analyst/parse-tolerant.d.ts
 /**
  * Forgiving pre-parse for analyst findings. Weak models routinely emit
@@ -749,5 +817,5 @@ declare function isProposalFinding(finding: unknown): finding is ProposalFinding
  */
 declare function assertProposalFindings(findings: unknown, context?: string): ReadonlyArray<ProposalFinding>;
 //#endregion
-export { AGENT_RX_UPSTREAM_REVISION, ANALYST_BENCHMARK_COST_LEDGER_FILE, ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM, ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES, ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256, ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256, ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256, ANALYST_BENCHMARK_EVIDENCE_PACKAGE_VERSION, ANALYST_BENCHMARK_HELP, ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM, ANALYST_BENCHMARK_IMPLEMENTATION_FILES, ANALYST_BENCHMARK_IMPLEMENTATION_SHA256, ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE, ANALYST_BENCHMARK_MANIFEST_FILE, ANALYST_BENCHMARK_OBSERVATIONS_FILE, ANALYST_SEVERITIES, type AgentRxBenchmarkCaseOptions, type AgentRxCalibrationRunnerSummary, type AgentRxCalibrationSummary, type AgentRxFailure, type AgentRxPrediction, type AgentRxPredictionReport, type AgentRxRow, type Analyst, type AnalystBenchmarkArtifact, type AnalystBenchmarkCase, type AnalystBenchmarkCommandConfig, type AnalystBenchmarkCommandDependencies, type AnalystBenchmarkDatasetRef, type AnalystBenchmarkDescriptor, type AnalystBenchmarkError, type AnalystBenchmarkLabelState, type AnalystBenchmarkLocalRunReceipt, type AnalystBenchmarkObservation, type AnalystBenchmarkOutput, type AnalystBenchmarkProgressRow, type AnalystBenchmarkProvenance, type AnalystBenchmarkResult, type AnalystBenchmarkRunIdentity, type AnalystBenchmarkRunManifest, type AnalystBenchmarkRunner, type AnalystBenchmarkSummary, type AnalystComparisonMetric, type AnalystContext, type AnalystCost, type AnalystEvidenceExpectation, type AnalystEvidenceResolution, type AnalystEvidenceResolutionError, type AnalystEvidenceResolver, type AnalystFinding, type AnalystFindingScore, type AnalystHooks, type AnalystInputKind, type AnalystIssueExpectation, type AnalystLatencyDistribution, type AnalystMetricComparison, AnalystRegistry, type AnalystRegistryOptions, type AnalystRequirements, type AnalystRunEvent, type AnalystRunInputs, type AnalystRunResult, type AnalystRunSummary, type AnalystRunnerComparison, type AnalystSeverity, type AnalystUsageReceipt, type BehavioralAnalystOptions, type BudgetPolicy, CODE_TRACE_BENCH_ANALYST_PROMPT, CONTROL_INTEGRITY_ANALYST, type ChatCallOpts, type ChatClient, type ChatRequest, type ChatResponse, type ChatTransport, type CliBridgeTransportOpts, type CodeTraceBenchCaseOptions, type CodeTraceBenchLabelOptions, type CodeTraceBenchLabelSet, type CodeTraceBenchRow, type CodeTraceBlockDiagnostics, type CodeTraceCalibrationRunnerSummary, type CodeTraceCalibrationSummary, type CodeTraceFailureBlock, type CodeTraceStageAnnotation, type CodeTracerLabelGroup, type CodeTracerPredictionAdapterOptions, type CodeTracerPredictions, type CodeTracerStepLabel, ControlIntegrityAnalyst, type CreateChatClientOpts, type CreateTraceAnalystOptions, type CustomTransportOpts, DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES, DEFAULT_TRACE_ANALYST_KINDS, DEFAULT_TRACE_ANALYST_LIMITS, type DefaultAnalystRegistryOptions, type DefineCustomAnalystOptions, type DefineExactCustomAnalystOptions, type DiffPolicy, type DirectProviderTransportOpts, type DspyRlmTraceEngineOptions, type EvidenceRef, type ExactAnalystBudgetPolicy, type ExactAnalystBudgetSnapshot, type ExactAnalystExecutionPlanSnapshot, type ExactAnalystRunCompletion, type ExactAnalystRunEvent, ExactAnalystRunExecutionError, type ExactAnalystRunPolicySnapshot, type ExactAnalystRunResult, type ExactAnalystRunSummary, type ExactAnalystSnapshot, type ExactCapableAnalyst, type ExactExecutionComponentIdentity, type ExactExecutionComponentSnapshot, type ExactRegistryRunOpts, FAILURE_MODE_KIND_SPEC, FINDING_SUBJECT_GRAMMAR_PROMPT, FINDING_SUBJECT_KINDS, FINDING_SUBJECT_SYNTAX, type FindingSubject, type FindingSubjectKind, FindingSubjectStringSchema, type FindingsDiff, FindingsStore, IMPROVEMENT_KIND_SPEC, type JudgeAdapterOpts, KIND_EXPECTED_SUBJECTS, KNOWLEDGE_GAP_KIND_SPEC, KNOWLEDGE_POISONING_KIND_SPEC, type LoadedVerificationArtifacts, MAX_INCORRECT_BLOCKS, MAX_INCORRECT_BLOCK_STEPS, type MockTransportOpts, type PersistedFinding, type PreparedPublicAnalystBenchmark, type ProposalFinding, type ProposalFindingOrigin, type PublicAnalystBenchmarkDataset, type PublicAnalystBenchmarkModelConfig, type PublicBenchmarkDistributions, type PublicBenchmarkSelectionReport, type PublicBenchmarkValueDistribution, RAW_FINDING_SCHEMA_PROMPT, type RawAnalystEvidence, RawAnalystEvidenceSchema, type RawAnalystFinding, RawAnalystFindingSchema, type RegistryRunOpts, type RouterTransportOpts, type RunAnalystBenchmarkOptions, type RunCriticAdapterOpts, SKILL_USAGE_ANALYST, type SandboxSdkTransportOpts, type SemanticConceptJudgeAdapterOpts, SkillUsageAnalyst, type SkillUsageRecord, type SkillUsageReport, type SkillUsageScanConfig, type StepLabelAdapterOptions, type TraceAnalysisEngine, type TraceAnalysisEngineRequest, type TraceAnalysisEngineResult, type TraceAnalystDefinition, type TraceAnalystLimits, type TraceToolGroupName, type UpstreamPredictionAdapterOptions, type VerificationArtifactFile, type VerificationArtifactManifest, type VerificationArtifactRole, type VerificationAvailabilitySummary, type VerificationOutcome, type VerificationOutcomeSource, type VerificationOutcomeStatus, type VerificationResultFile, type VerifierAdapterOpts, adaptPublicBenchmarkFindings, agentRxBenchmarkCase, agentRxPredictionsToFindings, analystBenchmarkDependencyLockDigest, analystBenchmarkImplementationDigest, appendVerificationArtifactsToOtlp, assertExactRegistryRunOpts, assertProposalFindings, behavioralAnalyst, buildDefaultAnalystRegistry, buildSkillUsageReport, buildTraceToolsForGroup, codeTraceBenchCase, codeTracerPredictionsToFindings, coerceJson, coerceToFindingRows, compareAnalystRunners, computeFindingId, createChatClient, createDspyRlmTraceEngine, createJudgeAdapter, createPublicBenchmarkDirectRunner, createPublicBenchmarkRlmRunner, createRunCriticAdapter, createSemanticConceptJudgeAdapter, createTraceAnalyst, createVerifierAdapter, defaultIsMaterial, defineCustomAnalyst, defineTraceAnalyst, deriveEfficiencyFindings, diffFindings, emitControlIntegrityFindings, emitSkillUsageFindings, emptyPublicBenchmarkRunner, evidenceRefsFromRawFinding, expandCodeTraceFailureBlocks, findingSubjectGrammarPromptFor, isProposalFinding, liftSeverity, loadCodeTraceVerificationArtifacts, loadPublicBenchmarkRows, makeFinding, makeProposalFinding, normalizeAgentRxCategory, normalizeBenchmarkLabel, parseFindingSubject, parseRawFinding, parseVerificationOutcome, preparePublicAnalystBenchmark, publicBenchmarkDistributions, publicBenchmarkProtocolSha256, publicBenchmarkRlmInstructions, publicBenchmarkSelectionReport, publicBenchmarkSystemPrompt, readAnalystBenchmarkArtifact, registryBenchmarkRunner, renderAgentRxCalibrationMarkdown, renderAnalystBenchmarkMarkdown, renderCodeTraceCalibrationMarkdown, renderFindingSubject, renderPriorFindings, renderUpstreamFindings, resolveTraceAnalystLimits, roundAgentRxStep, runAnalystBenchmark, runAnalystBenchmarkCommand, runTraceAnalyst, scoreAnalystFindings, selectPublicBenchmarkRows, stripCodeFences, summarizeAgentRxCalibration, summarizeCodeTraceCalibration, traceStoreEvidenceResolver };
+export { AGENT_RX_UPSTREAM_REVISION, ANALYST_BENCHMARK_COST_LEDGER_FILE, ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM, ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES, ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256, ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256, ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256, ANALYST_BENCHMARK_EVIDENCE_PACKAGE_VERSION, ANALYST_BENCHMARK_HELP, ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM, ANALYST_BENCHMARK_IMPLEMENTATION_FILES, ANALYST_BENCHMARK_IMPLEMENTATION_SHA256, ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE, ANALYST_BENCHMARK_MANIFEST_FILE, ANALYST_BENCHMARK_OBSERVATIONS_FILE, ANALYST_SEVERITIES, type AgentRxBenchmarkCaseOptions, type AgentRxCalibrationRunnerSummary, type AgentRxCalibrationSummary, type AgentRxFailure, type AgentRxPrediction, type AgentRxPredictionReport, type AgentRxRow, type Analyst, type AnalystBenchmarkArtifact, type AnalystBenchmarkCase, type AnalystBenchmarkCommandConfig, type AnalystBenchmarkCommandDependencies, type AnalystBenchmarkDatasetRef, type AnalystBenchmarkDescriptor, type AnalystBenchmarkError, type AnalystBenchmarkLabelState, type AnalystBenchmarkLocalRunReceipt, type AnalystBenchmarkObservation, type AnalystBenchmarkOutput, type AnalystBenchmarkProgressRow, type AnalystBenchmarkProvenance, type AnalystBenchmarkResult, type AnalystBenchmarkRunIdentity, type AnalystBenchmarkRunManifest, type AnalystBenchmarkRunner, type AnalystBenchmarkSummary, type AnalystComparisonMetric, type AnalystContext, type AnalystCost, type AnalystEvidenceExpectation, type AnalystEvidenceResolution, type AnalystEvidenceResolutionError, type AnalystEvidenceResolver, type AnalystFinding, type AnalystFindingScore, type AnalystHooks, type AnalystInputKind, type AnalystInstructionsOverride, type AnalystIssueExpectation, type AnalystLatencyDistribution, type AnalystMetricComparison, AnalystRegistry, type AnalystRegistryOptions, type AnalystRequirements, type AnalystRunEvent, type AnalystRunInputs, type AnalystRunResult, type AnalystRunSummary, type AnalystRunnerComparison, type AnalystSeverity, type AnalystUsageReceipt, type BehavioralAnalystOptions, type BudgetPolicy, CODE_TRACE_BENCH_ANALYST_PROMPT, CONTROL_INTEGRITY_ANALYST, type ChatCallOpts, type ChatClient, type ChatRequest, type ChatResponse, type ChatTransport, type CliBridgeTransportOpts, type CodeTraceBenchCaseOptions, type CodeTraceBenchLabelOptions, type CodeTraceBenchLabelSet, type CodeTraceBenchRow, type CodeTraceBlockDiagnostics, type CodeTraceCalibrationRunnerSummary, type CodeTraceCalibrationSummary, type CodeTraceFailureBlock, type CodeTraceStageAnnotation, type CodeTracerLabelGroup, type CodeTracerPredictionAdapterOptions, type CodeTracerPredictions, type CodeTracerStepLabel, ControlIntegrityAnalyst, type CreateChatClientOpts, type CreateTraceAnalystOptions, type CustomTransportOpts, DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES, DEFAULT_TRACE_ANALYST_KINDS, DEFAULT_TRACE_ANALYST_LIMITS, type DefaultAnalystRegistryOptions, type DefineCustomAnalystOptions, type DefineExactCustomAnalystOptions, type DiffPolicy, type DirectProviderTransportOpts, type DspyRlmTraceEngineOptions, type EvidenceRef, type ExactAnalystBudgetPolicy, type ExactAnalystBudgetSnapshot, type ExactAnalystExecutionPlanSnapshot, type ExactAnalystRunCompletion, type ExactAnalystRunEvent, ExactAnalystRunExecutionError, type ExactAnalystRunPolicySnapshot, type ExactAnalystRunResult, type ExactAnalystRunSummary, type ExactAnalystSnapshot, type ExactCapableAnalyst, type ExactExecutionComponentIdentity, type ExactExecutionComponentSnapshot, type ExactRegistryRunOpts, FAILURE_MODE_KIND_SPEC, FINDING_SUBJECT_GRAMMAR_PROMPT, FINDING_SUBJECT_KINDS, FINDING_SUBJECT_SYNTAX, type FindingSubject, type FindingSubjectKind, FindingSubjectStringSchema, type FindingsDiff, FindingsStore, IMPROVEMENT_KIND_SPEC, type JudgeAdapterOpts, KIND_EXPECTED_SUBJECTS, KNOWLEDGE_GAP_KIND_SPEC, KNOWLEDGE_POISONING_KIND_SPEC, type LoadedVerificationArtifacts, MAX_INCORRECT_BLOCKS, MAX_INCORRECT_BLOCK_STEPS, type MockTransportOpts, type PersistedFinding, type PreparedPublicAnalystBenchmark, type ProposalFinding, type ProposalFindingOrigin, type PublicAnalystBenchmarkDataset, type PublicAnalystBenchmarkModelConfig, type PublicBenchmarkDistributions, type PublicBenchmarkSelectionReport, type PublicBenchmarkValueDistribution, RAW_FINDING_SCHEMA_PROMPT, type RawAnalystEvidence, RawAnalystEvidenceSchema, type RawAnalystFinding, RawAnalystFindingSchema, type RegistryRunOpts, type RouterTransportOpts, type RunAnalystBenchmarkOptions, type RunCriticAdapterOpts, SKILL_USAGE_ANALYST, type SandboxSdkTransportOpts, type SemanticConceptJudgeAdapterOpts, SkillUsageAnalyst, type SkillUsageRecord, type SkillUsageReport, type SkillUsageScanConfig, type StepLabelAdapterOptions, type TraceAnalysisEngine, type TraceAnalysisEngineRequest, type TraceAnalysisEngineResult, type TraceAnalystDefinition, type TraceAnalystLimits, type TraceToolGroupName, type UpstreamPredictionAdapterOptions, type VerificationArtifactFile, type VerificationArtifactManifest, type VerificationArtifactRole, type VerificationAvailabilitySummary, type VerificationOutcome, type VerificationOutcomeSource, type VerificationOutcomeStatus, type VerificationResultFile, type VerifierAdapterOpts, adaptPublicBenchmarkFindings, agentRxBenchmarkCase, agentRxPredictionsToFindings, analystBenchmarkDependencyLockDigest, analystBenchmarkImplementationDigest, analystInstructionsOverrideFromText, appendVerificationArtifactsToOtlp, assertExactRegistryRunOpts, assertProposalFindings, behavioralAnalyst, buildDefaultAnalystRegistry, buildSkillUsageReport, buildTraceToolsForGroup, codeTraceBenchCase, codeTracerPredictionsToFindings, coerceJson, coerceToFindingRows, compareAnalystRunners, computeFindingId, createChatClient, createDspyRlmTraceEngine, createJudgeAdapter, createPublicBenchmarkDirectRunner, createPublicBenchmarkRlmRunner, createRunCriticAdapter, createSemanticConceptJudgeAdapter, createTraceAnalyst, createVerifierAdapter, defaultIsMaterial, defineCustomAnalyst, defineTraceAnalyst, deriveEfficiencyFindings, diffFindings, effectiveAnalystProtocolSha256, emitControlIntegrityFindings, emitSkillUsageFindings, emptyPublicBenchmarkRunner, evidenceRefsFromRawFinding, expandCodeTraceFailureBlocks, findingSubjectGrammarPromptFor, isProposalFinding, liftSeverity, loadCodeTraceVerificationArtifacts, loadPublicBenchmarkRows, makeFinding, makeProposalFinding, normalizeAgentRxCategory, normalizeBenchmarkLabel, parseFindingSubject, parseRawFinding, parseVerificationOutcome, preparePublicAnalystBenchmark, publicBenchmarkDistributions, publicBenchmarkProtocolSha256, publicBenchmarkRlmInstructions, publicBenchmarkSelectionReport, publicBenchmarkSystemPrompt, readAnalystBenchmarkArtifact, readAnalystInstructionsOverride, registryBenchmarkRunner, renderAgentRxCalibrationMarkdown, renderAnalystBenchmarkMarkdown, renderCodeTraceCalibrationMarkdown, renderFindingSubject, renderPriorFindings, renderUpstreamFindings, resolveTraceAnalystLimits, roundAgentRxStep, runAnalystBenchmark, runAnalystBenchmarkCommand, runTraceAnalyst, scoreAnalystFindings, selectPublicBenchmarkRows, stripCodeFences, summarizeAgentRxCalibration, summarizeAnalystBenchmarkRunner, summarizeCodeTraceCalibration, traceStoreEvidenceResolver };
 //# sourceMappingURL=index.d.ts.map

package/dist/analyst/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","names":[],"sources":["../../src/analyst/adapters.ts","../../src/analyst/benchmark-agentrx-calibration.ts","../../src/analyst/benchmark-dataset-types.ts","../../src/analyst/benchmark-dataset-agentrx.ts","../../src/analyst/benchmark-dataset-codetrace.ts","../../src/analyst/benchmark-dataset-utils.ts","../../src/analyst/benchmark-verification-outcome.ts","../../src/analyst/benchmark-verification-artifacts.ts","../../src/analyst/benchmark-public-types.ts","../../src/analyst/benchmark-public-adapters.ts","../../src/analyst/benchmark-public-data.ts","../../src/analyst/benchmark-public-prompt.ts","../../src/analyst/benchmark-public-model.ts","../../src/analyst/benchmark-public-rlm.ts","../../src/analyst/benchmark-comparison.ts","../../src/analyst/benchmark-public-calibration.ts","../../src/analyst/benchmark-command-artifact.ts","../../src/analyst/benchmark-command-result.ts","../../src/analyst/benchmark-command.ts","../../src/analyst/benchmark-implementation.ts","../../src/analyst/benchmark-report.ts","../../src/analyst/parse-tolerant.ts","../../src/analyst/proposal-findings.ts"],"mappings":";;;;;;;;;;;;iBA4CgB,aAAa,GAAG,WAAgB;UAe/B,oBAAoB;EACnC;EACA;EACA,UAAU,mBAAmB;;;;;EAK7B,UAAU,KAAK,cAAc;;iBAGf,sBAAsB,KAAK,MAAM,oBAAoB,OAAO,QAAQ;UAwEnE;EACf;EACA;EACA,SAAS;;EAET;;iBAGc,uBAAuB,OAAM,uBAA4B,QAAQ;UAiEhE;EACf;EACA;EACA,OAAO;;EAEP,MAAM;;EAEN,OAAO;;EAEP;;iBAGc,mBAAmB,MAAM,mBAAmB,QAAQ;UAiDnD;EACf;EACA;;EAEA,UAAU,KAAK;;EAEf;;iBAGc,kCACd,OAAM,kCACL,QAAQ;;;cC5RE;UAEI;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA,SAAS;;iBAGK,4BACd,QAAQ,wBACR,2BACC;iBAkBa,iCAAiC,SAAS;;;KCtD9C;UAEK;EACf,YAAY;EACZ;EACA;EACA;EACA;EACA;;UAGe;EACf,eAAe;EACf,mBAAmB;EACnB;IAAe,YAAY;IAAY;;EACvC,wBAAwB;EACxB;EACA;EACA;;UAGe;EACf,UAAU;EACV;EACA;EACA;EACA;;UAGe;EACf,UAAU;EACV,mBAAmB;EACnB;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA,iBAAiB;;KAGP,0CAEC,sCACA,iCACA;UAEI;EACf;EACA;EACA;EACA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA,oCAAoC;;UAGrB;EACf,eAAe;EACf,WAAW,sBAAsB;;KAGvB;UAEK;;;;;EAKf,WAAW;;UAGI,kCACP,yBACN;UAEa,oCAAoC;EACnD;;EAEA;;UAGe,yCAAyC;EACxD;EACA;EACA;EACA;;UAGe,2CACP,kCACN;;;iBCtGY,qBAAqB,QACnC,KAAK,YACL,OAAO,QACP,UAAS,8BACR,qBAAqB;;iBAoGR,6BACd,mBAAmB,YACnB,iBACA,UAAS,mCACR;iBA6Ea,yBAAyB;;iBAoNzB,iBAAiB;;;iBC7YjB,mBAAmB,QACjC,KAAK,mBACL,OAAO,QACP,UAAS,4BACR,qBAAqB;;iBAwFR,gCACd,2BACA,aAAa,uBACb,UAAS,qCACR;;;iBClHa,wBAAwB;;;KCA5B;UAEK;EACf;EACA;EACA,QAAQ;;UAGO;EACf,QAAQ;EACR;EAKA;IAAe;IAAe;;EAC9B,SAAS;EACT;EACA;EACA;EACA;;UAGe;EACf;EACA;;iBA8Ec,yBACd,gBAAgB,2BACf;;;cChGU;KAED;UAEK;EACf,MAAM;EACN;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,SAAS;EACT;EACA;EACA;EACA;EACA;EACA,OAAO;EACP,cAAc;EACd,UAAU,OAAO;;UAGF;EACf,UAAU;EACV,SAAS;EACT,OAAO,MAAM;IAA6B;;;iBAatB,mCAAmC;EACvD;EACA,KAAK;EACL;IACE,QAAQ;iBAwHI,kCACd,kBACA,iBACA,WAAW,6BACX;;;KChLU;UAEK;EACf;EACA;EACA;EACA;EACA;;EAEA,UAAU;;EAEV;EACA;IACE,SAAS;IACT;IACA;IACA;IACA;;EAEF,aAAa;EACb;IACE;IACA;;;EAGF,mBAAmB;;UAGJ;EACf,OAAO,qBAAqB;EAC5B;EACA;EACA;EACA,YAAY;IACV;IACA;IACA;;EAEF,uBAAuB;EACvB,WAAW;;UAGI;EACf;EACA;EACA,QAAQ;;UAGO;EACf,OAAO;EACP,OAAO;EACP,OAAO;EACP,YAAY;EACZ,QAAQ;;UAGO;EACf;EACA;EACA;EACA;EACA;EACA;EACA,QAAQ;EACR,UAAU;;;;;;;;;;UCnDK;EACf;EACA;;EAEA;EACA;EACA,UAAU;EACV;EACA;EACA;EACA;EACA,WAAW;;;;;;;;;UAUI;EACf;EACA;;EAEA,kCAAkC;;EAElC;;EAEA;;EAEA;;iBAGc,8BAA8B,uBAAuB;iBAiB/C,6BAA6B;EACjD,SAAS;EACT;EACA,mBAAmB;EACnB;EACA,OAAO;EACP,SAAS;IACP;EAAU,UAAU;EAAkB,aAAa;;;;;;;;;iBAqKjC,6BAA6B;EACjD;EACA,iBAAiB;EACjB,OAAO;EACP;EACA;EACA,SAAS;IACP;EAAU,UAAU;EAAkB,aAAa;;;;iBCzMjC,wBACpB,eACC,QAAQ,MAAM;iBA2BD,0BACd,SAAS,+BACT,eAAe,2BACf;EAAW;EAAe;IACzB,MAAM;iBAwBO,6BACd,SAAS,+BACT,eAAe,4BACd;iBAyCa,+BACd,SAAS,+BACT,iBAAiB,2BACjB,mBAAmB,2BACnB,eACC;iBAcmB,8BAA8B;EAClD,SAAS;EACT;EACA;EACA;EACA;EACA;EACA;IACE,QAAQ;;;;;;;cCxKC;;;;cAKA;cAMA;;iBA0FG,4BAA4B,SAAS;;iBAerC,+BAA+B,SAAS;;;;iBAaxC,8BAA8B,SAAS;;;;iBCzEvC,kCACd,SAAS,+BACT,QAAQ,oCACP,uBAAuB;;;;iBCtCV,+BACd,SAAS,+BACT,QAAQ,oCACP,uBAAuB;;;KC7Bd;UAoBK;EACf,QAAQ;EACR;;EAEA;;EAEA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,SAAS;;iBASK,sBACd,QAAQ,wBACR;EACE;EACA;EACA;EACA;EACA;IAED;;;UCnEc;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,SAAS;;iBAGK,8BACd,QAAQ,yBACP;iBAca,mCAAmC,SAAS;;;UCrC3C;EACf;EACA;EACA;IACE,SAAS;IACT;IACA;IACA;IACA;IACA,YAAY;MAAQ;MAAiB;MAAsB;;IAC3D,uBAAuB;IACvB,0BAA0B;IAC1B;MACE;MACA;MACA;MACA,QAAQ;;IAEV;MACE;MACA;MACA;MACA;MACA;MACA;MACA;MACA;MACA;MACA;;;EAGJ,QAAQ;EACR,aAAa;EACb,uBAAuB;EACvB,qBAAqB;;UAGN;EACf;EACA;EACA;EACA;IACE;IACA;IACA;;;UAIa;EACf;IACE,SAAS;IACT;IACA;IACA;MACE;MACA;MACA;;IAEF;IACA;IACA;IACA;IACA;IACA;IACA;IACA;IACA;IACA;;EAEF;IACE;IACA;IACA;IACA,YAAY;MAAQ;MAAiB;MAAsB;;IAC3D;IACA;;;UAIa;EACf;EACA;EACA;EACA;EACA,UAAU;;UAGK;EACf;EACA;EACA;EACA;IACE;IACA;IACA;IACA;IACA;IACA;;EAEF;EACA;IACE;IACA;IACA;;EAEF;IACE;IACA;IACA;IACA;IACA;IACA;;;UAIa;EACf;EACA;EACA;EACA,aAAa;EACb;;cAGW;cACA;cACA;cACA;;;iBCxHS,6BACpB,eACC,QAAQ;;;UC8DM;EACf,uBACE,SAAS,+BACT,QAAQ,sCACL,uBAAuB;;;;;;;;;;KAWlB;UAEK;EACf,SAAS;EACT,SAAS;EACT;EACA;EACA;EACA;EACA;EACA;EACA,OAAO;EACP;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;iBAKoB,2BACpB,yBACA,MAAK,OAAO,YACZ,eAAc,sCACb;cAsSU;;;cCraA;cAEA;cAEA;cAOA;;;;;;;;cAUA;cAEA;cAGA;cAGA;cAoFA;iBAGG;iBAIA;;;iBCrHA,+BACd,QAAQ,wBACR,uBAAsB;;;;;;;;;;;;;;;iBCQR,gBAAgB;;;;;iBAgBhB,WAAW;;;;;;;iBAeX,oBAAoB;;;;iBCXpB,kBAAkB,mBAAmB,WAAW;;;;;;iBAShD,uBACd,mBACA,mBACC,cAAc"}
+{"version":3,"file":"index.d.ts","names":[],"sources":["../../src/analyst/adapters.ts","../../src/analyst/benchmark-agentrx-calibration.ts","../../src/analyst/benchmark-dataset-types.ts","../../src/analyst/benchmark-dataset-agentrx.ts","../../src/analyst/benchmark-dataset-codetrace.ts","../../src/analyst/benchmark-dataset-utils.ts","../../src/analyst/benchmark-verification-outcome.ts","../../src/analyst/benchmark-verification-artifacts.ts","../../src/analyst/benchmark-public-types.ts","../../src/analyst/benchmark-public-adapters.ts","../../src/analyst/benchmark-public-data.ts","../../src/analyst/benchmark-public-prompt.ts","../../src/analyst/benchmark-public-model.ts","../../src/analyst/benchmark-public-rlm.ts","../../src/analyst/benchmark-comparison.ts","../../src/analyst/benchmark-public-calibration.ts","../../src/analyst/benchmark-command-artifact.ts","../../src/analyst/benchmark-command-result.ts","../../src/analyst/benchmark-command.ts","../../src/analyst/benchmark-implementation.ts","../../src/analyst/benchmark-instructions-override.ts","../../src/analyst/benchmark-report.ts","../../src/analyst/benchmark-summary.ts","../../src/analyst/parse-tolerant.ts","../../src/analyst/proposal-findings.ts"],"mappings":";;;;;;;;;;;;iBA4CgB,aAAa,GAAG,WAAgB;UAe/B,oBAAoB;EACnC;EACA;EACA,UAAU,mBAAmB;;;;;EAK7B,UAAU,KAAK,cAAc;;iBAGf,sBAAsB,KAAK,MAAM,oBAAoB,OAAO,QAAQ;UAwEnE;EACf;EACA;EACA,SAAS;;EAET;;iBAGc,uBAAuB,OAAM,uBAA4B,QAAQ;UAiEhE;EACf;EACA;EACA,OAAO;;EAEP,MAAM;;EAEN,OAAO;;EAEP;;iBAGc,mBAAmB,MAAM,mBAAmB,QAAQ;UAiDnD;EACf;EACA;;EAEA,UAAU,KAAK;;EAEf;;iBAGc,kCACd,OAAM,kCACL,QAAQ;;;cC5RE;UAEI;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA,SAAS;;iBAGK,4BACd,QAAQ,wBACR,2BACC;iBAkBa,iCAAiC,SAAS;;;KCtD9C;UAEK;EACf,YAAY;EACZ;EACA;EACA;EACA;EACA;;UAGe;EACf,eAAe;EACf,mBAAmB;EACnB;IAAe,YAAY;IAAY;;EACvC,wBAAwB;EACxB;EACA;EACA;;UAGe;EACf,UAAU;EACV;EACA;EACA;EACA;;UAGe;EACf,UAAU;EACV,mBAAmB;EACnB;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA,iBAAiB;;KAGP,0CAEC,sCACA,iCACA;UAEI;EACf;EACA;EACA;EACA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA,oCAAoC;;UAGrB;EACf,eAAe;EACf,WAAW,sBAAsB;;KAGvB;UAEK;;;;;EAKf,WAAW;;UAGI,kCACP,yBACN;UAEa,oCAAoC;EACnD;;EAEA;;UAGe,yCAAyC;EACxD;EACA;EACA;EACA;;UAGe,2CACP,kCACN;;;iBCtGY,qBAAqB,QACnC,KAAK,YACL,OAAO,QACP,UAAS,8BACR,qBAAqB;;iBAoGR,6BACd,mBAAmB,YACnB,iBACA,UAAS,mCACR;iBA6Ea,yBAAyB;;iBAoNzB,iBAAiB;;;iBC7YjB,mBAAmB,QACjC,KAAK,mBACL,OAAO,QACP,UAAS,4BACR,qBAAqB;;iBAwFR,gCACd,2BACA,aAAa,uBACb,UAAS,qCACR;;;iBClHa,wBAAwB;;;KCA5B;UAEK;EACf;EACA;EACA,QAAQ;;UAGO;EACf,QAAQ;EACR;EAKA;IAAe;IAAe;;EAC9B,SAAS;EACT;EACA;EACA;EACA;;UAGe;EACf;EACA;;iBA8Ec,yBACd,gBAAgB,2BACf;;;cChGU;KAED;UAEK;EACf,MAAM;EACN;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,SAAS;EACT;EACA;EACA;EACA;EACA;EACA,OAAO;EACP,cAAc;EACd,UAAU,OAAO;;UAGF;EACf,UAAU;EACV,SAAS;EACT,OAAO,MAAM;IAA6B;;;iBAatB,mCAAmC;EACvD;EACA,KAAK;EACL;IACE,QAAQ;iBAwHI,kCACd,kBACA,iBACA,WAAW,6BACX;;;KChLU;;;;;;;;UASK;;WAEN;;WAEA;;UAGM;EACf;EACA;EACA;EACA;EACA;;EAEA,UAAU;;EAEV;;EAEA,uBAAuB;EACvB;IACE,SAAS;IACT;IACA;IACA;IACA;;;;;;IAMA;;EAEF,aAAa;EACb;IACE;IACA;;;EAGF,mBAAmB;;UAGJ;EACf,OAAO,qBAAqB;EAC5B;EACA;EACA;EACA,YAAY;IACV;IACA;IACA;;EAEF,uBAAuB;EACvB,WAAW;;UAGI;EACf;EACA;EACA,QAAQ;;UAGO;EACf,OAAO;EACP,OAAO;EACP,OAAO;EACP,YAAY;EACZ,QAAQ;;UAGO;EACf;EACA;EACA;EACA;EACA;EACA;EACA,QAAQ;EACR,UAAU;;;;;;;;;;UCzEK;EACf;EACA;;EAEA;EACA;EACA,UAAU;EACV;EACA;EACA;EACA;EACA,WAAW;;;;;;;;;UAUI;EACf;EACA;;EAEA,kCAAkC;;EAElC;;EAEA;;EAEA;;EAEA;;EAEA;;;;;;;;UASe;EACf;EACA,OAAO;;iBAGO,8BAA8B,uBAAuB;iBAiB/C,6BAA6B;EACjD,SAAS;EACT;EACA,mBAAmB;EACnB;EACA,OAAO;EACP,SAAS;IACP;EACF,UAAU;EACV,aAAa;;EAEb,aAAa;;;;;;;;;iBA4MO,6BAA6B;EACjD;EACA,iBAAiB;EACjB,OAAO;EACP;EACA;EACA,SAAS;IACP;EACF,UAAU;EACV,aAAa;EACb,YAAY;;;;iBCtQQ,wBACpB,eACC,QAAQ,MAAM;iBA2BD,0BACd,SAAS,+BACT,eAAe,2BACf;EAAW;EAAe;IACzB,MAAM;iBAwBO,6BACd,SAAS,+BACT,eAAe,4BACd;iBAyCa,+BACd,SAAS,+BACT,iBAAiB,2BACjB,mBAAmB,2BACnB,eACC;iBAcmB,8BAA8B;EAClD,SAAS;EACT;EACA;EACA;EACA;EACA;EACA;IACE,QAAQ;;;;;;;cCxKC;;;;cAKA;cAMA;;iBAiGG,4BAA4B,SAAS;;iBAerC,+BAA+B,SAAS;;;;iBAaxC,8BAA8B,SAAS;;;;iBChFvC,kCACd,SAAS,+BACT,QAAQ,oCACP,uBAAuB;;;;iBChCV,+BACd,SAAS,+BACT,QAAQ,oCACP,uBAAuB;;;KCnCd;UAoBK;EACf,QAAQ;EACR;;EAEA;;EAEA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,SAAS;;iBASK,sBACd,QAAQ,wBACR;EACE;EACA;EACA;EACA;EACA;IAED;;;UCnEc;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,SAAS;;iBAGK,8BACd,QAAQ,yBACP;iBAca,mCAAmC,SAAS;;;UCrC3C;EACf;EACA;EACA;IACE,SAAS;IACT;IACA;IACA;IACA;IACA,YAAY;MAAQ;MAAiB;MAAsB;;IAC3D,uBAAuB;IACvB,0BAA0B;IAC1B;MACE;MACA;MACA;MACA,QAAQ;;IAEV;MACE;MACA;;MAEA;MACA;MACA;MACA;MACA;MACA;MACA;;MAEA;MACA;MACA;;;EAGJ,QAAQ;EACR,aAAa;EACb,uBAAuB;EACvB,qBAAqB;;UAGN;EACf;EACA;EACA;EACA;IACE;IACA;IACA;;;UAIa;EACf;IACE,SAAS;IACT;IACA;IACA;MACE;MACA;MACA;;IAEF;IACA;IACA;IACA;;IAEA;IACA;IACA;IACA;;IAEA;IACA;IACA;IACA;;EAEF;IACE;IACA;IACA;IACA,YAAY;MAAQ;MAAiB;MAAsB;;IAC3D;IACA;;;UAIa;EACf;EACA;EACA;EACA;EACA,UAAU;;UAGK;EACf;EACA;EACA;EACA;IACE;IACA;IACA;IACA;IACA;IACA;;EAEF;EACA;IACE;IACA;IACA;;EAEF;IACE;IACA;IACA;IACA;IACA;IACA;;;UAIa;EACf;EACA;EACA;EACA,aAAa;EACb;;cAGW;cACA;cACA;cACA;;;iBChIS,6BACpB,eACC,QAAQ;;;UCiEM;EACf,uBACE,SAAS,+BACT,QAAQ,sCACL,uBAAuB;;;;;;;;;;KAWlB;UAEK;EACf,SAAS;EACT,SAAS;EACT;EACA;EACA;EACA;EACA;EACA;EACA,OAAO;EACP;EACA;EACA;EACA;;EAEA;EACA;EACA;EACA;EACA;EACA;;iBAKoB,2BACpB,yBACA,MAAK,OAAO,YACZ,eAAc,sCACb;cAiTU;;;cCrbA;cAEA;cAEA;cAOA;;;;;;;;cAUA;cAEA;cAGA;cAGA;cAsFA;iBAGG;iBAIA;;;;iBCjHA,oCAAoC,eAAe;;iBAQnD,gCAAgC,eAAe;;;;;;;;;;;iBAyB/C,+BACd,SAAS,+BACT,WAAW,KAAK;;;iBCzCF,+BACd,QAAQ,wBACR,uBAAsB;;;iBCER,gCACd,kBACA,uBAAuB,gCACtB;;;;;;;;;;;;;;;iBCGa,gBAAgB;;;;;iBAgBhB,WAAW;;;;;;;iBAeX,oBAAoB;;;;iBCXpB,kBAAkB,mBAAmB,WAAW;;;;;;iBAShD,uBACd,mBACA,mBACC,cAAc"}

package/dist/analyst/index.js

@@ -1,12 +1,12 @@
 import { i as CostLedger } from "../cost-ledger-DMFxsLKr.js";
-import { C as createChatClient, _ as CONTROL_INTEGRITY_ANALYST, b as behavioralAnalyst, f as DEFAULT_TRACE_ANALYST_KINDS, g as FAILURE_MODE_KIND_SPEC, h as IMPROVEMENT_KIND_SPEC, i as assertExactRegistryRunOpts, m as KNOWLEDGE_GAP_KIND_SPEC, n as AnalystRegistry, p as KNOWLEDGE_POISONING_KIND_SPEC, r as ExactAnalystRunExecutionError, t as buildDefaultAnalystRegistry, v as ControlIntegrityAnalyst, x as deriveEfficiencyFindings, y as emitControlIntegrityFindings } from "../default-registry-BOiBXNFh.js";
-import { A as coerceJson, B as renderFindingSubject, D as RawAnalystFindingSchema, E as RawAnalystEvidenceSchema, F as FINDING_SUBJECT_SYNTAX, G as resolveTraceAnalystLimits, I as FindingSubjectStringSchema, L as KIND_EXPECTED_SUBJECTS, M as stripCodeFences, N as FINDING_SUBJECT_GRAMMAR_PROMPT, O as evidenceRefsFromRawFinding, P as FINDING_SUBJECT_KINDS, R as findingSubjectGrammarPromptFor, T as RAW_FINDING_SCHEMA_PROMPT, W as DEFAULT_TRACE_ANALYST_LIMITS, a as buildTraceToolsForGroup, i as runTraceAnalyst, j as coerceToFindingRows, k as parseRawFinding, n as renderPriorFindings, r as renderUpstreamFindings, t as createTraceAnalyst, w as ANALYST_SEVERITIES, z as parseFindingSubject } from "../kind-factory-DB7nIs35.js";
+import { C as createChatClient, _ as CONTROL_INTEGRITY_ANALYST, b as behavioralAnalyst, f as DEFAULT_TRACE_ANALYST_KINDS, g as FAILURE_MODE_KIND_SPEC, h as IMPROVEMENT_KIND_SPEC, i as assertExactRegistryRunOpts, m as KNOWLEDGE_GAP_KIND_SPEC, n as AnalystRegistry, p as KNOWLEDGE_POISONING_KIND_SPEC, r as ExactAnalystRunExecutionError, t as buildDefaultAnalystRegistry, v as ControlIntegrityAnalyst, x as deriveEfficiencyFindings, y as emitControlIntegrityFindings } from "../default-registry-BupFSXSN.js";
+import { A as coerceJson, B as renderFindingSubject, D as RawAnalystFindingSchema, E as RawAnalystEvidenceSchema, F as FINDING_SUBJECT_SYNTAX, G as resolveTraceAnalystLimits, I as FindingSubjectStringSchema, L as KIND_EXPECTED_SUBJECTS, M as stripCodeFences, N as FINDING_SUBJECT_GRAMMAR_PROMPT, O as evidenceRefsFromRawFinding, P as FINDING_SUBJECT_KINDS, R as findingSubjectGrammarPromptFor, T as RAW_FINDING_SCHEMA_PROMPT, W as DEFAULT_TRACE_ANALYST_LIMITS, a as buildTraceToolsForGroup, i as runTraceAnalyst, j as coerceToFindingRows, k as parseRawFinding, n as renderPriorFindings, r as renderUpstreamFindings, t as createTraceAnalyst, w as ANALYST_SEVERITIES, z as parseFindingSubject } from "../kind-factory-Bvwe3pup.js";
 import { a as computeFindingId, i as validateUsageSettlementTimeout, n as settleUsageReceiptFromCostLedger, o as makeFinding, s as makeProposalFinding } from "../usage-receipt-CgxMEBZq.js";
 import { n as isProposalFinding, t as assertProposalFindings } from "../proposal-findings-2GIUo1et.js";
 import { a as RunCritic, c as buildSkillUsageReport, d as defaultIsMaterial, f as diffFindings, h as defineTraceAnalyst, i as runSemanticConceptJudge, l as emitSkillUsageFindings, m as defineCustomAnalyst, n as SEMANTIC_CONCEPT_JUDGE_VERSION, o as SKILL_USAGE_ANALYST, s as SkillUsageAnalyst, u as FindingsStore } from "../semantic-concept-judge-DJQtFr95.js";
-import { t as createDspyRlmTraceEngine } from "../dspy-rlm-engine-Bkak4nzo.js";
-import { a as scoreAnalystFindings, n as runAnalystBenchmark, r as traceStoreEvidenceResolver, t as registryBenchmarkRunner } from "../benchmark-CYtcIF2V.js";
-import { A as ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256, B as ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE, C as publicBenchmarkSystemPrompt, D as parseVerificationOutcome, E as loadCodeTraceVerificationArtifacts, F as ANALYST_BENCHMARK_IMPLEMENTATION_FILES, G as summarizeAgentRxCalibration, H as ANALYST_BENCHMARK_OBSERVATIONS_FILE, I as ANALYST_BENCHMARK_IMPLEMENTATION_SHA256, J as agentRxBenchmarkCase, K as codeTraceBenchCase, L as analystBenchmarkDependencyLockDigest, M as ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256, N as ANALYST_BENCHMARK_EVIDENCE_PACKAGE_VERSION, O as ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM, P as ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM, Q as normalizeBenchmarkLabel, R as analystBenchmarkImplementationDigest, S as publicBenchmarkRlmInstructions, T as appendVerificationArtifactsToOtlp, U as AGENT_RX_UPSTREAM_REVISION, V as ANALYST_BENCHMARK_MANIFEST_FILE, W as renderAgentRxCalibrationMarkdown, X as normalizeAgentRxCategory, Y as agentRxPredictionsToFindings, Z as roundAgentRxStep, _ as expandCodeTraceFailureBlocks, a as renderCodeTraceCalibrationMarkdown, b as MAX_INCORRECT_BLOCK_STEPS, c as createPublicBenchmarkRlmRunner, d as preparePublicAnalystBenchmark, f as publicBenchmarkDistributions, g as emptyPublicBenchmarkRunner, h as adaptPublicBenchmarkFindings, i as readAnalystBenchmarkArtifact, j as ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256, k as ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES, l as createPublicBenchmarkDirectRunner, m as selectPublicBenchmarkRows, n as runAnalystBenchmarkCommand, o as summarizeCodeTraceCalibration, p as publicBenchmarkSelectionReport, q as codeTracerPredictionsToFindings, r as renderAnalystBenchmarkMarkdown, s as compareAnalystRunners, t as ANALYST_BENCHMARK_HELP, u as loadPublicBenchmarkRows, v as CODE_TRACE_BENCH_ANALYST_PROMPT, w as DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES, x as publicBenchmarkProtocolSha256, y as MAX_INCORRECT_BLOCKS, z as ANALYST_BENCHMARK_COST_LEDGER_FILE } from "../benchmark-command-Bo_hkhQf.js";
+import { t as createDspyRlmTraceEngine } from "../dspy-rlm-engine-Bw8Zz9CX.js";
+import { a as scoreAnalystFindings, i as summarizeAnalystBenchmarkRunner, n as runAnalystBenchmark, r as traceStoreEvidenceResolver, t as registryBenchmarkRunner } from "../benchmark-CYtcIF2V.js";
+import { $ as normalizeAgentRxCategory, A as parseVerificationOutcome, B as analystBenchmarkDependencyLockDigest, C as MAX_INCORRECT_BLOCK_STEPS, D as DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES, E as publicBenchmarkSystemPrompt, F as ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256, G as ANALYST_BENCHMARK_OBSERVATIONS_FILE, H as ANALYST_BENCHMARK_COST_LEDGER_FILE, I as ANALYST_BENCHMARK_EVIDENCE_PACKAGE_VERSION, J as summarizeAgentRxCalibration, K as AGENT_RX_UPSTREAM_REVISION, L as ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM, M as ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES, N as ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256, O as appendVerificationArtifactsToOtlp, P as ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256, Q as agentRxPredictionsToFindings, R as ANALYST_BENCHMARK_IMPLEMENTATION_FILES, S as MAX_INCORRECT_BLOCKS, T as publicBenchmarkRlmInstructions, U as ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE, V as analystBenchmarkImplementationDigest, W as ANALYST_BENCHMARK_MANIFEST_FILE, X as codeTracerPredictionsToFindings, Y as codeTraceBenchCase, Z as agentRxBenchmarkCase, _ as compareAnalystRunners, a as preparePublicAnalystBenchmark, b as readAnalystInstructionsOverride, c as selectPublicBenchmarkRows, d as adaptPublicBenchmarkFindings, et as roundAgentRxStep, f as emptyPublicBenchmarkRunner, g as summarizeCodeTraceCalibration, h as renderCodeTraceCalibrationMarkdown, i as loadPublicBenchmarkRows, j as ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM, k as loadCodeTraceVerificationArtifacts, l as createPublicBenchmarkRlmRunner, m as readAnalystBenchmarkArtifact, n as runAnalystBenchmarkCommand, o as publicBenchmarkDistributions, p as expandCodeTraceFailureBlocks, q as renderAgentRxCalibrationMarkdown, r as renderAnalystBenchmarkMarkdown, s as publicBenchmarkSelectionReport, t as ANALYST_BENCHMARK_HELP, tt as normalizeBenchmarkLabel, u as createPublicBenchmarkDirectRunner, v as analystInstructionsOverrideFromText, w as publicBenchmarkProtocolSha256, x as CODE_TRACE_BENCH_ANALYST_PROMPT, y as effectiveAnalystProtocolSha256, z as ANALYST_BENCHMARK_IMPLEMENTATION_SHA256 } from "../benchmark-command-D8xXNVVh.js";
 //#region src/analyst/adapters.ts
 /**
 * Adapter factories — lift each existing agent-eval primitive into the
@@ -294,6 +294,6 @@ function createSemanticConceptJudgeAdapter(opts = {}) {
 	};
 }
 //#endregion
-export { AGENT_RX_UPSTREAM_REVISION, ANALYST_BENCHMARK_COST_LEDGER_FILE, ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM, ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES, ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256, ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256, ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256, ANALYST_BENCHMARK_EVIDENCE_PACKAGE_VERSION, ANALYST_BENCHMARK_HELP, ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM, ANALYST_BENCHMARK_IMPLEMENTATION_FILES, ANALYST_BENCHMARK_IMPLEMENTATION_SHA256, ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE, ANALYST_BENCHMARK_MANIFEST_FILE, ANALYST_BENCHMARK_OBSERVATIONS_FILE, ANALYST_SEVERITIES, AnalystRegistry, CODE_TRACE_BENCH_ANALYST_PROMPT, CONTROL_INTEGRITY_ANALYST, ControlIntegrityAnalyst, DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES, DEFAULT_TRACE_ANALYST_KINDS, DEFAULT_TRACE_ANALYST_LIMITS, ExactAnalystRunExecutionError, FAILURE_MODE_KIND_SPEC, FINDING_SUBJECT_GRAMMAR_PROMPT, FINDING_SUBJECT_KINDS, FINDING_SUBJECT_SYNTAX, FindingSubjectStringSchema, FindingsStore, IMPROVEMENT_KIND_SPEC, KIND_EXPECTED_SUBJECTS, KNOWLEDGE_GAP_KIND_SPEC, KNOWLEDGE_POISONING_KIND_SPEC, MAX_INCORRECT_BLOCKS, MAX_INCORRECT_BLOCK_STEPS, RAW_FINDING_SCHEMA_PROMPT, RawAnalystEvidenceSchema, RawAnalystFindingSchema, SKILL_USAGE_ANALYST, SkillUsageAnalyst, adaptPublicBenchmarkFindings, agentRxBenchmarkCase, agentRxPredictionsToFindings, analystBenchmarkDependencyLockDigest, analystBenchmarkImplementationDigest, appendVerificationArtifactsToOtlp, assertExactRegistryRunOpts, assertProposalFindings, behavioralAnalyst, buildDefaultAnalystRegistry, buildSkillUsageReport, buildTraceToolsForGroup, codeTraceBenchCase, codeTracerPredictionsToFindings, coerceJson, coerceToFindingRows, compareAnalystRunners, computeFindingId, createChatClient, createDspyRlmTraceEngine, createJudgeAdapter, createPublicBenchmarkDirectRunner, createPublicBenchmarkRlmRunner, createRunCriticAdapter, createSemanticConceptJudgeAdapter, createTraceAnalyst, createVerifierAdapter, defaultIsMaterial, defineCustomAnalyst, defineTraceAnalyst, deriveEfficiencyFindings, diffFindings, emitControlIntegrityFindings, emitSkillUsageFindings, emptyPublicBenchmarkRunner, evidenceRefsFromRawFinding, expandCodeTraceFailureBlocks, findingSubjectGrammarPromptFor, isProposalFinding, liftSeverity, loadCodeTraceVerificationArtifacts, loadPublicBenchmarkRows, makeFinding, makeProposalFinding, normalizeAgentRxCategory, normalizeBenchmarkLabel, parseFindingSubject, parseRawFinding, parseVerificationOutcome, preparePublicAnalystBenchmark, publicBenchmarkDistributions, publicBenchmarkProtocolSha256, publicBenchmarkRlmInstructions, publicBenchmarkSelectionReport, publicBenchmarkSystemPrompt, readAnalystBenchmarkArtifact, registryBenchmarkRunner, renderAgentRxCalibrationMarkdown, renderAnalystBenchmarkMarkdown, renderCodeTraceCalibrationMarkdown, renderFindingSubject, renderPriorFindings, renderUpstreamFindings, resolveTraceAnalystLimits, roundAgentRxStep, runAnalystBenchmark, runAnalystBenchmarkCommand, runTraceAnalyst, scoreAnalystFindings, selectPublicBenchmarkRows, stripCodeFences, summarizeAgentRxCalibration, summarizeCodeTraceCalibration, traceStoreEvidenceResolver };
+export { AGENT_RX_UPSTREAM_REVISION, ANALYST_BENCHMARK_COST_LEDGER_FILE, ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM, ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES, ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256, ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256, ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256, ANALYST_BENCHMARK_EVIDENCE_PACKAGE_VERSION, ANALYST_BENCHMARK_HELP, ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM, ANALYST_BENCHMARK_IMPLEMENTATION_FILES, ANALYST_BENCHMARK_IMPLEMENTATION_SHA256, ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE, ANALYST_BENCHMARK_MANIFEST_FILE, ANALYST_BENCHMARK_OBSERVATIONS_FILE, ANALYST_SEVERITIES, AnalystRegistry, CODE_TRACE_BENCH_ANALYST_PROMPT, CONTROL_INTEGRITY_ANALYST, ControlIntegrityAnalyst, DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES, DEFAULT_TRACE_ANALYST_KINDS, DEFAULT_TRACE_ANALYST_LIMITS, ExactAnalystRunExecutionError, FAILURE_MODE_KIND_SPEC, FINDING_SUBJECT_GRAMMAR_PROMPT, FINDING_SUBJECT_KINDS, FINDING_SUBJECT_SYNTAX, FindingSubjectStringSchema, FindingsStore, IMPROVEMENT_KIND_SPEC, KIND_EXPECTED_SUBJECTS, KNOWLEDGE_GAP_KIND_SPEC, KNOWLEDGE_POISONING_KIND_SPEC, MAX_INCORRECT_BLOCKS, MAX_INCORRECT_BLOCK_STEPS, RAW_FINDING_SCHEMA_PROMPT, RawAnalystEvidenceSchema, RawAnalystFindingSchema, SKILL_USAGE_ANALYST, SkillUsageAnalyst, adaptPublicBenchmarkFindings, agentRxBenchmarkCase, agentRxPredictionsToFindings, analystBenchmarkDependencyLockDigest, analystBenchmarkImplementationDigest, analystInstructionsOverrideFromText, appendVerificationArtifactsToOtlp, assertExactRegistryRunOpts, assertProposalFindings, behavioralAnalyst, buildDefaultAnalystRegistry, buildSkillUsageReport, buildTraceToolsForGroup, codeTraceBenchCase, codeTracerPredictionsToFindings, coerceJson, coerceToFindingRows, compareAnalystRunners, computeFindingId, createChatClient, createDspyRlmTraceEngine, createJudgeAdapter, createPublicBenchmarkDirectRunner, createPublicBenchmarkRlmRunner, createRunCriticAdapter, createSemanticConceptJudgeAdapter, createTraceAnalyst, createVerifierAdapter, defaultIsMaterial, defineCustomAnalyst, defineTraceAnalyst, deriveEfficiencyFindings, diffFindings, effectiveAnalystProtocolSha256, emitControlIntegrityFindings, emitSkillUsageFindings, emptyPublicBenchmarkRunner, evidenceRefsFromRawFinding, expandCodeTraceFailureBlocks, findingSubjectGrammarPromptFor, isProposalFinding, liftSeverity, loadCodeTraceVerificationArtifacts, loadPublicBenchmarkRows, makeFinding, makeProposalFinding, normalizeAgentRxCategory, normalizeBenchmarkLabel, parseFindingSubject, parseRawFinding, parseVerificationOutcome, preparePublicAnalystBenchmark, publicBenchmarkDistributions, publicBenchmarkProtocolSha256, publicBenchmarkRlmInstructions, publicBenchmarkSelectionReport, publicBenchmarkSystemPrompt, readAnalystBenchmarkArtifact, readAnalystInstructionsOverride, registryBenchmarkRunner, renderAgentRxCalibrationMarkdown, renderAnalystBenchmarkMarkdown, renderCodeTraceCalibrationMarkdown, renderFindingSubject, renderPriorFindings, renderUpstreamFindings, resolveTraceAnalystLimits, roundAgentRxStep, runAnalystBenchmark, runAnalystBenchmarkCommand, runTraceAnalyst, scoreAnalystFindings, selectPublicBenchmarkRows, stripCodeFences, summarizeAgentRxCalibration, summarizeAnalystBenchmarkRunner, summarizeCodeTraceCalibration, traceStoreEvidenceResolver };
 
 //# sourceMappingURL=index.js.map