@tangle-network/agent-eval 0.50.2 → 0.51.0

package/CHANGELOG.md

@@ -4,6 +4,26 @@ All notable changes to `@tangle-network/agent-eval` and its sibling `agent-eval-
 
 ---
 
+## [0.51.0] — 2026-05-27 — skillOptDriver (SkillOpt methodology as a substrate driver)
+
+### Added
+
+- **`skillOptDriver`** in `/campaign`. A section-aware, bounded-edit `ImprovementDriver` for structured natural-language procedures (SKILL.md files, runbooks, sectioned system prompts, judge rubrics with dimensions). Implements the SkillOpt methodology (Microsoft, 2026): treat the skill document as a trainable optimization target, train the procedure not the weights, constrain each generation to ≤N targeted edits to prevent useful-rule overwrites.
+  - **Edit-budget enforcement** — candidates that exceed `editBudget * 2` sentence-level diffs vs baseline are rejected at parse time. SkillOpt's "edit budget functions as a textual learning rate."
+  - **Section preservation** — H2 headings (or an explicit `preserveSections` allowlist) MUST appear unchanged in every candidate. Candidates that delete or rename sections are rejected.
+  - **Surface-typed** — throws on non-string surfaces; agent-runtime's `improvementDriver` handles code-tier.
+- `extractH2Sections(text)` + `countSentenceEdits(a, b)` exported as named helpers for consumers writing custom drivers with similar invariants.
+
+### Scope (honest)
+
+This is **batch SkillOpt** — one LLM call per generation produces all N candidates with the budget enforced as a prompt instruction + post-parse rejection. **Per-edit iteration** (propose 1 edit → validate → accept-or-reject → propose next) is a future 0.52.0 enhancement that needs a new `IncrementalImprovementDriver` interface; the substrate's current batch `ImprovementDriver` can run SkillOpt-style behavior with `populationSize=1` + `maxGenerations=N`, but a single driver invocation can't iterate per-edit yet. Tracked.
+
+### Notes
+
+Selectable alongside `gepaDriver` and `evolutionaryDriver`. Use when the surface IS a structured doc; use `gepaDriver` when the surface is unstructured prose.
+
+---
+
 ## [0.50.2] — 2026-05-27 — actionability fixes from real-data dogfood
 
 ### Added

package/dist/campaign/index.d.ts

@@ -1,13 +1,72 @@
 export { C as CampaignStorage, D as DefaultProductionGateOptions, E as EvolutionaryDriverOptions, G as GepaDriverOptions, H as HeldOutGateOptions, O as OpenAutoPrOptions, a as OpenAutoPrResult, R as RunCampaignOptions, b as RunEvalOptions, c as RunImprovementLoopOptions, d as RunImprovementLoopResult, e as RunOptimizationOptions, f as RunOptimizationResult, g as composeGate, h as defaultProductionGate, i as evolutionaryDriver, j as fsCampaignStorage, k as gepaDriver, l as heldOutGate, m as inMemoryCampaignStorage, o as openAutoPr, r as runCampaign, n as runEval, p as runImprovementLoop, q as runOptimization, s as surfaceHash } from '../run-improvement-loop-BPMjNKMJ.js';
-import { L as LabeledScenarioStore, c as LabeledScenarioWrite, d as LabeledScenarioSampleArgs, e as LabeledScenarioRecord, C as CodeSurface } from '../types-Dbj5gu8n.js';
-export { f as CampaignAggregates, g as CampaignArtifactWriter, h as CampaignCellResult, i as CampaignCostMeter, j as CampaignResult, k as CampaignTraceWriter, b as DispatchContext, D as DispatchFn, G as Gate, l as GateContext, m as GateDecision, n as GateResult, o as GenerationCandidate, p as GenerationRecord, I as ImprovementDriver, q as JudgeAggregate, a as JudgeConfig, r as JudgeDimension, J as JudgeScore, s as LabeledScenarioSource, M as MutableSurface, t as Mutator, O as OptimizerConfig, P as ProposeContext, R as RedactionStatus, S as Scenario, u as ScenarioAggregate, v as SessionScript, T as TraceSpan } from '../types-Dbj5gu8n.js';
-import '../llm-client-BXVRUZyX.js';
-import '../errors-mje_cKOs.js';
-import '../raw-provider-sink-C46HDghv.js';
+import { L as LlmClientOptions } from '../llm-client-BXVRUZyX.js';
+import { I as ImprovementDriver, L as LabeledScenarioStore, c as LabeledScenarioWrite, d as LabeledScenarioSampleArgs, e as LabeledScenarioRecord, C as CodeSurface } from '../types-Dbj5gu8n.js';
+export { f as CampaignAggregates, g as CampaignArtifactWriter, h as CampaignCellResult, i as CampaignCostMeter, j as CampaignResult, k as CampaignTraceWriter, b as DispatchContext, D as DispatchFn, G as Gate, l as GateContext, m as GateDecision, n as GateResult, o as GenerationCandidate, p as GenerationRecord, q as JudgeAggregate, a as JudgeConfig, r as JudgeDimension, J as JudgeScore, s as LabeledScenarioSource, M as MutableSurface, t as Mutator, O as OptimizerConfig, P as ProposeContext, R as RedactionStatus, S as Scenario, u as ScenarioAggregate, v as SessionScript, T as TraceSpan } from '../types-Dbj5gu8n.js';
 import '../red-team-30II1T4o.js';
 import '../dataset-BlwAtYYf.js';
+import '../errors-mje_cKOs.js';
 import '../store-Db2Bv8Cf.js';
 import '../run-record-BGY6bHRh.js';
+import '../raw-provider-sink-C46HDghv.js';
+
+/**
+ * @experimental
+ *
+ * `skillOptDriver` — a section-aware, bounded-edit `ImprovementDriver` for
+ * structured natural-language procedures (SKILL.md files, runbooks, sectioned
+ * system prompts, judge rubrics with dimensions). Implements the SkillOpt
+ * methodology (Microsoft, 2026): treat the skill document as a trainable
+ * optimization target, train the procedure not the weights, constrain each
+ * generation to ≤N targeted edits to prevent useful-rule overwrites.
+ *
+ * Differs from `gepaDriver` in two specific ways:
+ *
+ * 1. **Bounded edits.** Each candidate must differ from the baseline by at
+ *    most `editBudget` sentence-level changes. The "edit budget functions
+ *    as a textual learning rate" — without it, an LLM proposal can rewrite
+ *    so much that useful prior rules get overwritten.
+ *
+ * 2. **Section preservation.** When the surface is a structured doc, the
+ *    H2 headers (and an opt-in `preserveSections` allowlist) are
+ *    load-bearing for discoverability. Candidates that delete or rename
+ *    preserved sections are rejected at parse time.
+ *
+ * Selectable alongside `gepaDriver` and `evolutionaryDriver`. Use this when
+ * the surface IS a structured doc; use `gepaDriver` when the surface is
+ * unstructured prose.
+ */
+
+interface SkillOptDriverOptions {
+    /** Router transport (apiKey/baseUrl). */
+    llm: LlmClientOptions;
+    /** Model that performs the reflection. */
+    model: string;
+    /** What is being optimized — appears in the reflection prompt for orientation. */
+    target: string;
+    /** Max edits per generation — SkillOpt's "textual learning rate".
+     *  Default 3. Lower = more conservative, higher = more exploratory. */
+    editBudget?: number;
+    /** Section headings the driver MUST preserve. When the surface is a
+     *  structured skill doc, sections are load-bearing for discoverability.
+     *  Default: auto-detected from H2 headers in the baseline. */
+    preserveSections?: string[];
+    /** Surface-specific mutation levers offered to the model. */
+    mutationPrimitives?: string[];
+    /** Top/bottom scenarios surfaced as evidence each generation. Default 3. */
+    evidenceK?: number;
+    /** Reflection sampling temperature. Default 0.7. */
+    temperature?: number;
+    /** Reflection max tokens. Default 6000. */
+    maxTokens?: number;
+}
+/** Internal — exported for tests. */
+declare function extractH2Sections(text: string): string[];
+/** Sentence-level edit distance. Counts distinct sentence add/remove/replace
+ *  ops between baseline and candidate using a normalised line-by-line diff.
+ *  Imperfect (treats trivial whitespace as identical) but tight enough to
+ *  bound an LLM rewrite. Exported for tests. */
+declare function countSentenceEdits(baseline: string, candidate: string): number;
+declare function skillOptDriver(opts: SkillOptDriverOptions): ImprovementDriver;
 
 /**
  * @experimental
@@ -124,4 +183,4 @@ declare function gitWorktreeAdapter(opts: GitWorktreeAdapterOptions): WorktreeAd
  *  as a ref under the adapter's worktree dir. */
 declare function resolveWorktreePath(surface: CodeSurface, worktreeDir?: string): string;
 
-export { CodeSurface, FsLabeledScenarioStore, type FsLabeledScenarioStoreOptions, type GitWorktreeAdapterOptions, LabeledScenarioRecord, LabeledScenarioSampleArgs, LabeledScenarioStore, LabeledScenarioStoreError, LabeledScenarioWrite, type Worktree, type WorktreeAdapter, WorktreeAdapterError, gitWorktreeAdapter, resolveWorktreePath };
+export { CodeSurface, FsLabeledScenarioStore, type FsLabeledScenarioStoreOptions, type GitWorktreeAdapterOptions, ImprovementDriver, LabeledScenarioRecord, LabeledScenarioSampleArgs, LabeledScenarioStore, LabeledScenarioStoreError, LabeledScenarioWrite, type SkillOptDriverOptions, type Worktree, type WorktreeAdapter, WorktreeAdapterError, countSentenceEdits, extractH2Sections, gitWorktreeAdapter, resolveWorktreePath, skillOptDriver };

package/dist/campaign/index.js

@@ -15,15 +15,133 @@ import {
   inMemoryCampaignStorage,
   runCampaign
 } from "../chunk-J3EIOI3O.js";
-import "../chunk-N4SBKEPJ.js";
+import {
+  buildReflectionPrompt,
+  parseReflectionResponse
+} from "../chunk-N4SBKEPJ.js";
 import "../chunk-YV7J7X5N.js";
 import "../chunk-WP7SY7AI.js";
 import "../chunk-GGE4NNQT.js";
-import "../chunk-VXNVVBZO.js";
+import {
+  callLlm
+} from "../chunk-VXNVVBZO.js";
 import "../chunk-PC4UYEBM.js";
 import "../chunk-QYJT52YW.js";
 import "../chunk-NSBPE2FW.js";
 
+// src/campaign/drivers/skillopt.ts
+var REFLECTION_SYSTEM = 'You are an expert prompt engineer applying the SkillOpt methodology. You will edit a structured natural-language procedure under TWO HARD CONSTRAINTS: (1) preserve every H2 section heading verbatim \u2014 do NOT delete, rename, or merge sections; (2) make at most EDIT_BUDGET targeted sentence-level edits per candidate \u2014 bounded edits prevent overwriting useful prior rules. Output ONLY a JSON object of shape {"proposals":[{"label":string,"rationale":string,"payload":string}]} where each `payload` is the FULL improved skill text. No prose outside the JSON.';
+function extractH2Sections(text) {
+  const out = [];
+  for (const line of text.split("\n")) {
+    const match = /^##\s+(.+?)\s*$/.exec(line);
+    if (match) out.push(match[1]);
+  }
+  return out;
+}
+function countSentenceEdits(baseline, candidate) {
+  const norm = (s) => s.split(/(?<=[.!?])\s+|\n/g).map((p) => p.trim()).filter((p) => p.length > 0);
+  const a = new Set(norm(baseline));
+  const b = new Set(norm(candidate));
+  let edits = 0;
+  for (const s of a) if (!b.has(s)) edits++;
+  for (const s of b) if (!a.has(s)) edits++;
+  return edits;
+}
+function skillOptDriver(opts) {
+  const evidenceK = opts.evidenceK ?? 3;
+  const editBudget = opts.editBudget ?? 3;
+  if (editBudget < 1) {
+    throw new Error(
+      `skillOptDriver: editBudget must be >= 1, got ${editBudget} (use evolutionaryDriver with a noop mutator for measure-only runs)`
+    );
+  }
+  return {
+    kind: "skillopt",
+    async propose(ctx) {
+      if (typeof ctx.currentSurface !== "string") {
+        throw new Error(
+          `skillOptDriver: surface must be a string skill document; got ${typeof ctx.currentSurface}. Use evolutionaryDriver with a CodeSurface mutator for code-tier surfaces.`
+        );
+      }
+      const baseline = ctx.currentSurface;
+      const preserveSections = opts.preserveSections ?? extractH2Sections(baseline);
+      const { top, bottom, target } = buildEvidence(ctx, evidenceK, opts.target);
+      const reflectionUser = buildReflectionPrompt({
+        target,
+        parentPayload: baseline,
+        topTrials: top,
+        bottomTrials: bottom,
+        childCount: ctx.populationSize,
+        mutationPrimitives: opts.mutationPrimitives
+      });
+      const constraintPreamble = [
+        "",
+        "## SkillOpt constraints (hard rules \u2014 violations rejected)",
+        "",
+        `- Edit budget: at most ${editBudget} sentence-level edits per candidate.`,
+        "- Section preservation: every H2 heading below must appear unchanged in your output.",
+        ...preserveSections.map((s) => `  - \`## ${s}\``),
+        "",
+        "Reject any candidate in your own thinking that would delete a section, rename a heading, or exceed the edit budget. Make TARGETED, surgical edits \u2014 not rewrites.",
+        ""
+      ].join("\n");
+      const userPrompt = `${reflectionUser}${constraintPreamble}`;
+      const system = REFLECTION_SYSTEM.replace("EDIT_BUDGET", String(editBudget));
+      const result = await callLlm(
+        {
+          model: opts.model,
+          messages: [
+            { role: "system", content: system },
+            { role: "user", content: userPrompt }
+          ],
+          jsonMode: true,
+          temperature: opts.temperature ?? 0.7,
+          maxTokens: opts.maxTokens ?? 6e3
+        },
+        opts.llm
+      );
+      const proposals = parseReflectionResponse(result.content, ctx.populationSize);
+      const out = [];
+      for (const proposal of proposals) {
+        const text = typeof proposal.payload === "string" ? proposal.payload.trim() : "";
+        if (!text || text === baseline) continue;
+        if (!validateSections(text, preserveSections)) continue;
+        if (countSentenceEdits(baseline, text) > editBudget * 2) continue;
+        if (out.includes(text)) continue;
+        out.push(text);
+      }
+      return out;
+    }
+  };
+}
+function validateSections(candidate, required) {
+  if (required.length === 0) return true;
+  const have = new Set(extractH2Sections(candidate));
+  for (const section of required) {
+    if (!have.has(section)) return false;
+  }
+  return true;
+}
+function buildEvidence(ctx, evidenceK, baseTarget) {
+  const last = ctx.history.at(-1);
+  if (!last || last.candidates.length === 0) {
+    return { top: [], bottom: [], target: baseTarget };
+  }
+  const best = [...last.candidates].sort((a, b) => b.composite - a.composite)[0];
+  if (!best) return { top: [], bottom: [], target: baseTarget };
+  const byScore = [...best.scenarios].sort((a, b) => b.composite - a.composite);
+  const toTrace = (s) => ({
+    id: s.scenarioId,
+    score: s.composite
+  });
+  const top = byScore.slice(0, evidenceK).map(toTrace);
+  const bottom = byScore.slice(-evidenceK).reverse().map(toTrace);
+  const weakest = Object.entries(best.dimensions).sort((a, b) => a[1] - b[1]).slice(0, 3).map(([dim, value]) => `${dim} (${value.toFixed(2)})`);
+  const target = weakest.length > 0 ? `${baseTarget} \u2014 weakest dimensions: ${weakest.join(", ")}` : baseTarget;
+  return { top, bottom, target };
+}
+
 // src/campaign/labeled-store/fs-adapter.ts
 import { createHash } from "crypto";
 import { existsSync, mkdirSync, readFileSync, writeFileSync } from "fs";
@@ -283,8 +401,10 @@ export {
   LabeledScenarioStoreError,
   WorktreeAdapterError,
   composeGate,
+  countSentenceEdits,
   defaultProductionGate,
   evolutionaryDriver,
+  extractH2Sections,
   fsCampaignStorage,
   gepaDriver,
   gitWorktreeAdapter,
@@ -296,6 +416,7 @@ export {
   runEval,
   runImprovementLoop,
   runOptimization,
+  skillOptDriver,
   surfaceHash
 };
 //# sourceMappingURL=index.js.map

package/dist/campaign/index.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/campaign/labeled-store/fs-adapter.ts","../../src/campaign/worktree/index.ts"],"sourcesContent":["/**\n * @experimental\n *\n * Filesystem `LabeledScenarioStore` adapter. The default capture sink for\n * traces + eval artifacts. Production deployments typically swap for a\n * Turso/SQLite adapter (same interface).\n *\n * Records land as one JSONL file per source under `<root>/<source>.jsonl`.\n * Each line is a `LabeledScenarioRecord`. Append-only — no in-place edits.\n *\n * Safety properties enforced at write-time:\n *\n *   - **Provenance required**: writes without `source`, `sourceVersionHash`,\n *     `capturedAt`, `redactionStatus` are rejected. Closes the alignment\n *     reviewer's data-poisoning gap.\n *   - **Per-source rate limits**: optional `rateLimitBucket` + `maxWritesPerMinute`\n *     stops a single tenant/source from flooding the store.\n *\n * Safety properties enforced at sample-time:\n *\n *   - **Required split + capturedBefore**: substrate refuses to sample without\n *     an explicit `split` ('train' | 'test') AND a temporal cutoff. Eliminates\n *     accidental train/test contamination.\n *   - **Default training-source filter**: when the store is sampled with\n *     `split: 'train'`, production-trace records are EXCLUDED unless the\n *     caller passes `filter.source: 'production-trace'` explicitly. Closes\n *     the contamination-by-default gap flagged by the senior eval engineer.\n */\n\nimport { createHash } from 'node:crypto'\nimport { existsSync, mkdirSync, readFileSync, writeFileSync } from 'node:fs'\nimport { join } from 'node:path'\nimport type {\n  LabeledScenarioRecord,\n  LabeledScenarioSampleArgs,\n  LabeledScenarioSource,\n  LabeledScenarioStore,\n  LabeledScenarioWrite,\n} from '../types'\n\nexport interface FsLabeledScenarioStoreOptions {\n  /** Root directory for JSONL files. Created if missing. */\n  root: string\n  /** Per-source rate limit. When set, writes exceeding the cap are rejected\n   *  with a typed error. Default: no limit. */\n  maxWritesPerMinutePerBucket?: number\n  /** Test seam — override `Date.now()` for deterministic tests. */\n  now?: () => number\n}\n\nexport class LabeledScenarioStoreError extends Error {\n  constructor(\n    public readonly code: string,\n    message: string,\n  ) {\n    super(message)\n    this.name = 'LabeledScenarioStoreError'\n  }\n}\n\ninterface RateLimitState {\n  bucket: string\n  windowStartMs: number\n  count: number\n}\n\nexport class FsLabeledScenarioStore implements LabeledScenarioStore {\n  private readonly now: () => number\n  private readonly rateLimits = new Map<string, RateLimitState>()\n\n  constructor(private readonly options: FsLabeledScenarioStoreOptions) {\n    if (!existsSync(options.root)) mkdirSync(options.root, { recursive: true })\n    this.now = options.now ?? Date.now\n  }\n\n  async observe(write: LabeledScenarioWrite): Promise<void> {\n    this.assertProvenance(write)\n    this.assertRateLimit(write)\n    const record = this.toRecord(write)\n    const path = this.pathForSource(write.source)\n    const line = `${JSON.stringify(record)}\\n`\n    // Append atomically. For high-throughput a writev-friendly buffered\n    // implementation lands in the Turso adapter; FS adapter is for tests +\n    // local dev + small workloads.\n    appendLine(path, line)\n  }\n\n  async sample(args: LabeledScenarioSampleArgs): Promise<LabeledScenarioRecord[]> {\n    if (!args.split) {\n      throw new LabeledScenarioStoreError(\n        'split_required',\n        'sample() requires an explicit `split` (train | test) — substrate refuses ambiguous reads',\n      )\n    }\n    if (!args.capturedBefore) {\n      throw new LabeledScenarioStoreError(\n        'capturedBefore_required',\n        'sample() requires an explicit `capturedBefore` timestamp for temporal-split discipline',\n      )\n    }\n\n    const all: LabeledScenarioRecord[] = []\n    for (const source of ALL_SOURCES) {\n      // Default training-source filter: when sampling train, EXCLUDE\n      // production-trace records unless the caller asks for them.\n      if (args.split === 'train' && source === 'production-trace') {\n        const explicit = sourceFilterContains(args.filter?.source, 'production-trace')\n        if (!explicit) continue\n      }\n      const path = this.pathForSource(source)\n      if (!existsSync(path)) continue\n      const lines = readFileSync(path, 'utf8').split('\\n').filter(Boolean)\n      for (const line of lines) {\n        let record: LabeledScenarioRecord\n        try {\n          record = JSON.parse(line) as LabeledScenarioRecord\n        } catch {\n          continue\n        }\n        if (!matchesFilter(record, args, source)) continue\n        all.push(record)\n      }\n    }\n\n    // Deterministic order: by capturedAt ascending, then recordHash.\n    all.sort((a, b) => {\n      if (a.capturedAt !== b.capturedAt) return a.capturedAt.localeCompare(b.capturedAt)\n      return a.recordHash.localeCompare(b.recordHash)\n    })\n\n    return all.slice(0, args.count)\n  }\n\n  async size(): Promise<{ train: number; test: number; bySource: Record<string, number> }> {\n    const bySource: Record<string, number> = {}\n    let total = 0\n    for (const source of ALL_SOURCES) {\n      const path = this.pathForSource(source)\n      if (!existsSync(path)) {\n        bySource[source] = 0\n        continue\n      }\n      const count = readFileSync(path, 'utf8').split('\\n').filter(Boolean).length\n      bySource[source] = count\n      total += count\n    }\n    // FS adapter doesn't track split assignments per-record (split is\n    // computed at sample-time based on `capturedBefore`). For size(), we\n    // report `train`+`test` as the same total — split is a sampling concept.\n    return { train: total, test: total, bySource }\n  }\n\n  private assertProvenance(write: LabeledScenarioWrite): void {\n    if (!write.source) {\n      throw new LabeledScenarioStoreError(\n        'missing_source',\n        'LabeledScenarioWrite requires `source`',\n      )\n    }\n    if (!write.sourceVersionHash || write.sourceVersionHash.length === 0) {\n      throw new LabeledScenarioStoreError(\n        'missing_source_version',\n        'LabeledScenarioWrite requires `sourceVersionHash` (git sha or substrate version)',\n      )\n    }\n    if (!write.capturedAt) {\n      throw new LabeledScenarioStoreError(\n        'missing_captured_at',\n        'LabeledScenarioWrite requires `capturedAt` ISO timestamp',\n      )\n    }\n    if (!write.redactionStatus) {\n      throw new LabeledScenarioStoreError(\n        'missing_redaction_status',\n        'LabeledScenarioWrite requires explicit `redactionStatus` — raw / redacted-pii / redacted-secrets / fully-redacted',\n      )\n    }\n    if (!ALL_SOURCES.includes(write.source)) {\n      throw new LabeledScenarioStoreError(\n        'unknown_source',\n        `LabeledScenarioWrite.source must be one of: ${ALL_SOURCES.join(', ')}`,\n      )\n    }\n  }\n\n  private assertRateLimit(write: LabeledScenarioWrite): void {\n    const cap = this.options.maxWritesPerMinutePerBucket\n    if (!cap || !write.rateLimitBucket) return\n    const now = this.now()\n    const windowMs = 60_000\n    let state = this.rateLimits.get(write.rateLimitBucket)\n    if (!state || now - state.windowStartMs >= windowMs) {\n      state = { bucket: write.rateLimitBucket, windowStartMs: now, count: 0 }\n      this.rateLimits.set(write.rateLimitBucket, state)\n    }\n    if (state.count >= cap) {\n      throw new LabeledScenarioStoreError(\n        'rate_limit_exceeded',\n        `LabeledScenarioStore: bucket ${write.rateLimitBucket} exceeded ${cap} writes/min`,\n      )\n    }\n    state.count += 1\n  }\n\n  private toRecord(write: LabeledScenarioWrite): LabeledScenarioRecord {\n    const recordHash = sha256(\n      JSON.stringify({\n        id: write.scenario.id,\n        src: write.source,\n        at: write.capturedAt,\n        ver: write.sourceVersionHash,\n      }),\n    )\n    // FS adapter assigns split at sample-time, but we cache a hint here\n    // based on capturedAt vs the world's \"now\" — sampler overrides this.\n    return {\n      ...write,\n      recordHash,\n      split: 'train',\n    }\n  }\n\n  private pathForSource(source: string): string {\n    return join(this.options.root, `${source}.jsonl`)\n  }\n}\n\nconst ALL_SOURCES: LabeledScenarioWrite['source'][] = [\n  'production-trace',\n  'eval-run',\n  'manual',\n  'red-team',\n  'synthetic',\n]\n\nfunction sourceFilterContains(\n  filter: LabeledScenarioSource | LabeledScenarioSource[] | undefined,\n  needle: LabeledScenarioSource,\n): boolean {\n  if (!filter) return false\n  if (Array.isArray(filter)) return filter.includes(needle)\n  return filter === needle\n}\n\nfunction matchesFilter(\n  record: LabeledScenarioRecord,\n  args: LabeledScenarioSampleArgs,\n  source: string,\n): boolean {\n  // Temporal cutoff — train must be capturedAt < capturedBefore.\n  if (args.split === 'train' && record.capturedAt >= args.capturedBefore) return false\n  if (args.split === 'test' && record.capturedAt < args.capturedBefore) return false\n\n  const f = args.filter\n  if (!f) return true\n  if (f.kind && record.scenario.kind !== f.kind) return false\n  if (f.source) {\n    const sources = Array.isArray(f.source) ? f.source : [f.source]\n    if (!sources.includes(source as never)) return false\n  }\n  if (f.minComposite !== undefined || f.maxComposite !== undefined) {\n    const composites = Object.values(record.judgeScores).map((s) => s.composite)\n    const max = composites.length === 0 ? 0 : Math.max(...composites)\n    if (f.minComposite !== undefined && max < f.minComposite) return false\n    if (f.maxComposite !== undefined && max > f.maxComposite) return false\n  }\n  return true\n}\n\nfunction sha256(input: string): string {\n  return createHash('sha256').update(input).digest('hex').slice(0, 16)\n}\n\nfunction appendLine(path: string, line: string): void {\n  if (existsSync(path)) {\n    const existing = readFileSync(path, 'utf8')\n    writeFileSync(path, existing + line)\n  } else {\n    writeFileSync(path, line)\n  }\n}\n","/**\n * @experimental\n *\n * VCS-pluggable worktree adapter. One improvement = one worktree, PR-like\n * (multiple commits allowed). A code-tier driver's `propose()` creates a\n * worktree, an agent commits the change into it, and `finalize()` returns a\n * `CodeSurface{ worktreeRef }` the measurement checks out to run the worker\n * against the changed code. On promotion the worktree becomes the PR branch.\n *\n * The interface is VCS-agnostic so a future `jj` ([jj-vcs](https://github.com/jj-vcs/jj))\n * adapter can slot in without touching driver code. Only the git adapter\n * ships today. See `docs/design/self-improvement-engine.md`.\n */\n\nimport { execFileSync } from 'node:child_process'\nimport { existsSync } from 'node:fs'\nimport { basename, isAbsolute, join } from 'node:path'\nimport type { CodeSurface } from '../types'\n\nexport interface Worktree {\n  /** Absolute path to the checked-out worktree directory. */\n  path: string\n  /** The branch the worktree is on (becomes the PR branch on promotion). */\n  branch: string\n  /** The ref the worktree was forked from. */\n  baseRef: string\n}\n\nexport interface WorktreeAdapter {\n  /** Create an isolated worktree on a fresh branch off `baseRef`. */\n  create(opts: { baseRef: string; label: string }): Promise<Worktree>\n  /** Commit any pending changes in the worktree, then return a CodeSurface\n   *  pointing at it. The agent has already written its change into\n   *  `worktree.path` by the time this is called. */\n  finalize(worktree: Worktree, summary: string): Promise<CodeSurface>\n  /** Remove the worktree (and its branch) — called for losing candidates. */\n  discard(worktree: Worktree): Promise<void>\n}\n\nexport class WorktreeAdapterError extends Error {\n  constructor(\n    message: string,\n    readonly cause?: unknown,\n  ) {\n    super(message)\n    this.name = 'WorktreeAdapterError'\n  }\n}\n\nexport interface GitWorktreeAdapterOptions {\n  /** Repo root the worktrees fork from. */\n  repoRoot: string\n  /** Directory worktrees are created under. Default: `<repoRoot>/.worktrees`. */\n  worktreeDir?: string\n  /** Branch-name prefix. Default: `improve`. */\n  branchPrefix?: string\n  /** Test seam — defaults to a real `git` runner. */\n  git?: (args: string[], cwd: string) => string\n}\n\nfunction defaultGit(args: string[], cwd: string): string {\n  try {\n    return execFileSync('git', args, { cwd, encoding: 'utf8' }).trim()\n  } catch (err) {\n    const stderr =\n      err && typeof err === 'object' && 'stderr' in err\n        ? String((err as { stderr: unknown }).stderr)\n        : ''\n    throw new WorktreeAdapterError(`git ${args.join(' ')} failed: ${stderr || String(err)}`, err)\n  }\n}\n\n/** Slugify a label into a branch-safe segment. */\nfunction slug(label: string): string {\n  return (\n    label\n      .toLowerCase()\n      .replace(/[^a-z0-9]+/g, '-')\n      .replace(/^-+|-+$/g, '')\n      .slice(0, 48) || 'candidate'\n  )\n}\n\nexport function gitWorktreeAdapter(opts: GitWorktreeAdapterOptions): WorktreeAdapter {\n  const git = opts.git ?? defaultGit\n  const worktreeDir = opts.worktreeDir ?? join(opts.repoRoot, '.worktrees')\n  const branchPrefix = opts.branchPrefix ?? 'improve'\n\n  return {\n    async create({ baseRef, label }) {\n      const id = `${slug(label)}-${Date.now().toString(36)}-${Math.random().toString(36).slice(2, 6)}`\n      const branch = `${branchPrefix}/${id}`\n      const path = join(worktreeDir, id)\n      git(['worktree', 'add', '-b', branch, path, baseRef], opts.repoRoot)\n      return { path, branch, baseRef }\n    },\n\n    async finalize(worktree, summary) {\n      // Stage + commit any pending changes the agent left in the worktree.\n      // A no-op commit is refused by git, so only commit when the tree is dirty.\n      const status = git(['status', '--porcelain'], worktree.path)\n      if (status.length > 0) {\n        git(['add', '-A'], worktree.path)\n        git(['commit', '-m', summary], worktree.path)\n      }\n      return {\n        kind: 'code',\n        worktreeRef: worktree.path,\n        baseRef: worktree.baseRef,\n        summary,\n      }\n    },\n\n    async discard(worktree) {\n      // Remove the worktree, then delete its branch. Force-remove because the\n      // worktree may hold uncommitted experiment state we're discarding.\n      git(['worktree', 'remove', '--force', worktree.path], opts.repoRoot)\n      git(['branch', '-D', worktree.branch], opts.repoRoot)\n    },\n  }\n}\n\n/** Resolve a `CodeSurface`'s worktreeRef to a directory the measurement can\n *  run the worker in. A path ref is returned as-is; anything else is treated\n *  as a ref under the adapter's worktree dir. */\nexport function resolveWorktreePath(surface: CodeSurface, worktreeDir?: string): string {\n  if (isAbsolute(surface.worktreeRef) && existsSync(surface.worktreeRef)) return surface.worktreeRef\n  if (worktreeDir) return join(worktreeDir, basename(surface.worktreeRef))\n  return surface.worktreeRef\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;AA6BA,SAAS,kBAAkB;AAC3B,SAAS,YAAY,WAAW,cAAc,qBAAqB;AACnE,SAAS,YAAY;AAmBd,IAAM,4BAAN,cAAwC,MAAM;AAAA,EACnD,YACkB,MAChB,SACA;AACA,UAAM,OAAO;AAHG;AAIhB,SAAK,OAAO;AAAA,EACd;AAAA,EALkB;AAMpB;AAQO,IAAM,yBAAN,MAA6D;AAAA,EAIlE,YAA6B,SAAwC;AAAxC;AAC3B,QAAI,CAAC,WAAW,QAAQ,IAAI,EAAG,WAAU,QAAQ,MAAM,EAAE,WAAW,KAAK,CAAC;AAC1E,SAAK,MAAM,QAAQ,OAAO,KAAK;AAAA,EACjC;AAAA,EAH6B;AAAA,EAHZ;AAAA,EACA,aAAa,oBAAI,IAA4B;AAAA,EAO9D,MAAM,QAAQ,OAA4C;AACxD,SAAK,iBAAiB,KAAK;AAC3B,SAAK,gBAAgB,KAAK;AAC1B,UAAM,SAAS,KAAK,SAAS,KAAK;AAClC,UAAM,OAAO,KAAK,cAAc,MAAM,MAAM;AAC5C,UAAM,OAAO,GAAG,KAAK,UAAU,MAAM,CAAC;AAAA;AAItC,eAAW,MAAM,IAAI;AAAA,EACvB;AAAA,EAEA,MAAM,OAAO,MAAmE;AAC9E,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,KAAK,gBAAgB;AACxB,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AAEA,UAAM,MAA+B,CAAC;AACtC,eAAW,UAAU,aAAa;AAGhC,UAAI,KAAK,UAAU,WAAW,WAAW,oBAAoB;AAC3D,cAAM,WAAW,qBAAqB,KAAK,QAAQ,QAAQ,kBAAkB;AAC7E,YAAI,CAAC,SAAU;AAAA,MACjB;AACA,YAAM,OAAO,KAAK,cAAc,MAAM;AACtC,UAAI,CAAC,WAAW,IAAI,EAAG;AACvB,YAAM,QAAQ,aAAa,MAAM,MAAM,EAAE,MAAM,IAAI,EAAE,OAAO,OAAO;AACnE,iBAAW,QAAQ,OAAO;AACxB,YAAI;AACJ,YAAI;AACF,mBAAS,KAAK,MAAM,IAAI;AAAA,QAC1B,QAAQ;AACN;AAAA,QACF;AACA,YAAI,CAAC,cAAc,QAAQ,MAAM,MAAM,EAAG;AAC1C,YAAI,KAAK,MAAM;AAAA,MACjB;AAAA,IACF;AAGA,QAAI,KAAK,CAAC,GAAG,MAAM;AACjB,UAAI,EAAE,eAAe,EAAE,WAAY,QAAO,EAAE,WAAW,cAAc,EAAE,UAAU;AACjF,aAAO,EAAE,WAAW,cAAc,EAAE,UAAU;AAAA,IAChD,CAAC;AAED,WAAO,IAAI,MAAM,GAAG,KAAK,KAAK;AAAA,EAChC;AAAA,EAEA,MAAM,OAAmF;AACvF,UAAM,WAAmC,CAAC;AAC1C,QAAI,QAAQ;AACZ,eAAW,UAAU,aAAa;AAChC,YAAM,OAAO,KAAK,cAAc,MAAM;AACtC,UAAI,CAAC,WAAW,IAAI,GAAG;AACrB,iBAAS,MAAM,IAAI;AACnB;AAAA,MACF;AACA,YAAM,QAAQ,aAAa,MAAM,MAAM,EAAE,MAAM,IAAI,EAAE,OAAO,OAAO,EAAE;AACrE,eAAS,MAAM,IAAI;AACnB,eAAS;AAAA,IACX;AAIA,WAAO,EAAE,OAAO,OAAO,MAAM,OAAO,SAAS;AAAA,EAC/C;AAAA,EAEQ,iBAAiB,OAAmC;AAC1D,QAAI,CAAC,MAAM,QAAQ;AACjB,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,MAAM,qBAAqB,MAAM,kBAAkB,WAAW,GAAG;AACpE,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,MAAM,YAAY;AACrB,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,MAAM,iBAAiB;AAC1B,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,YAAY,SAAS,MAAM,MAAM,GAAG;AACvC,YAAM,IAAI;AAAA,QACR;AAAA,QACA,+CAA+C,YAAY,KAAK,IAAI,CAAC;AAAA,MACvE;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,gBAAgB,OAAmC;AACzD,UAAM,MAAM,KAAK,QAAQ;AACzB,QAAI,CAAC,OAAO,CAAC,MAAM,gBAAiB;AACpC,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,WAAW;AACjB,QAAI,QAAQ,KAAK,WAAW,IAAI,MAAM,eAAe;AACrD,QAAI,CAAC,SAAS,MAAM,MAAM,iBAAiB,UAAU;AACnD,cAAQ,EAAE,QAAQ,MAAM,iBAAiB,eAAe,KAAK,OAAO,EAAE;AACtE,WAAK,WAAW,IAAI,MAAM,iBAAiB,KAAK;AAAA,IAClD;AACA,QAAI,MAAM,SAAS,KAAK;AACtB,YAAM,IAAI;AAAA,QACR;AAAA,QACA,gCAAgC,MAAM,eAAe,aAAa,GAAG;AAAA,MACvE;AAAA,IACF;AACA,UAAM,SAAS;AAAA,EACjB;AAAA,EAEQ,SAAS,OAAoD;AACnE,UAAM,aAAa;AAAA,MACjB,KAAK,UAAU;AAAA,QACb,IAAI,MAAM,SAAS;AAAA,QACnB,KAAK,MAAM;AAAA,QACX,IAAI,MAAM;AAAA,QACV,KAAK,MAAM;AAAA,MACb,CAAC;AAAA,IACH;AAGA,WAAO;AAAA,MACL,GAAG;AAAA,MACH;AAAA,MACA,OAAO;AAAA,IACT;AAAA,EACF;AAAA,EAEQ,cAAc,QAAwB;AAC5C,WAAO,KAAK,KAAK,QAAQ,MAAM,GAAG,MAAM,QAAQ;AAAA,EAClD;AACF;AAEA,IAAM,cAAgD;AAAA,EACpD;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAEA,SAAS,qBACP,QACA,QACS;AACT,MAAI,CAAC,OAAQ,QAAO;AACpB,MAAI,MAAM,QAAQ,MAAM,EAAG,QAAO,OAAO,SAAS,MAAM;AACxD,SAAO,WAAW;AACpB;AAEA,SAAS,cACP,QACA,MACA,QACS;AAET,MAAI,KAAK,UAAU,WAAW,OAAO,cAAc,KAAK,eAAgB,QAAO;AAC/E,MAAI,KAAK,UAAU,UAAU,OAAO,aAAa,KAAK,eAAgB,QAAO;AAE7E,QAAM,IAAI,KAAK;AACf,MAAI,CAAC,EAAG,QAAO;AACf,MAAI,EAAE,QAAQ,OAAO,SAAS,SAAS,EAAE,KAAM,QAAO;AACtD,MAAI,EAAE,QAAQ;AACZ,UAAM,UAAU,MAAM,QAAQ,EAAE,MAAM,IAAI,EAAE,SAAS,CAAC,EAAE,MAAM;AAC9D,QAAI,CAAC,QAAQ,SAAS,MAAe,EAAG,QAAO;AAAA,EACjD;AACA,MAAI,EAAE,iBAAiB,UAAa,EAAE,iBAAiB,QAAW;AAChE,UAAM,aAAa,OAAO,OAAO,OAAO,WAAW,EAAE,IAAI,CAAC,MAAM,EAAE,SAAS;AAC3E,UAAM,MAAM,WAAW,WAAW,IAAI,IAAI,KAAK,IAAI,GAAG,UAAU;AAChE,QAAI,EAAE,iBAAiB,UAAa,MAAM,EAAE,aAAc,QAAO;AACjE,QAAI,EAAE,iBAAiB,UAAa,MAAM,EAAE,aAAc,QAAO;AAAA,EACnE;AACA,SAAO;AACT;AAEA,SAAS,OAAO,OAAuB;AACrC,SAAO,WAAW,QAAQ,EAAE,OAAO,KAAK,EAAE,OAAO,KAAK,EAAE,MAAM,GAAG,EAAE;AACrE;AAEA,SAAS,WAAW,MAAc,MAAoB;AACpD,MAAI,WAAW,IAAI,GAAG;AACpB,UAAM,WAAW,aAAa,MAAM,MAAM;AAC1C,kBAAc,MAAM,WAAW,IAAI;AAAA,EACrC,OAAO;AACL,kBAAc,MAAM,IAAI;AAAA,EAC1B;AACF;;;AC1QA,SAAS,oBAAoB;AAC7B,SAAS,cAAAA,mBAAkB;AAC3B,SAAS,UAAU,YAAY,QAAAC,aAAY;AAuBpC,IAAM,uBAAN,cAAmC,MAAM;AAAA,EAC9C,YACE,SACS,OACT;AACA,UAAM,OAAO;AAFJ;AAGT,SAAK,OAAO;AAAA,EACd;AAAA,EAJW;AAKb;AAaA,SAAS,WAAW,MAAgB,KAAqB;AACvD,MAAI;AACF,WAAO,aAAa,OAAO,MAAM,EAAE,KAAK,UAAU,OAAO,CAAC,EAAE,KAAK;AAAA,EACnE,SAAS,KAAK;AACZ,UAAM,SACJ,OAAO,OAAO,QAAQ,YAAY,YAAY,MAC1C,OAAQ,IAA4B,MAAM,IAC1C;AACN,UAAM,IAAI,qBAAqB,OAAO,KAAK,KAAK,GAAG,CAAC,YAAY,UAAU,OAAO,GAAG,CAAC,IAAI,GAAG;AAAA,EAC9F;AACF;AAGA,SAAS,KAAK,OAAuB;AACnC,SACE,MACG,YAAY,EACZ,QAAQ,eAAe,GAAG,EAC1B,QAAQ,YAAY,EAAE,EACtB,MAAM,GAAG,EAAE,KAAK;AAEvB;AAEO,SAAS,mBAAmB,MAAkD;AACnF,QAAM,MAAM,KAAK,OAAO;AACxB,QAAM,cAAc,KAAK,eAAeA,MAAK,KAAK,UAAU,YAAY;AACxE,QAAM,eAAe,KAAK,gBAAgB;AAE1C,SAAO;AAAA,IACL,MAAM,OAAO,EAAE,SAAS,MAAM,GAAG;AAC/B,YAAM,KAAK,GAAG,KAAK,KAAK,CAAC,IAAI,KAAK,IAAI,EAAE,SAAS,EAAE,CAAC,IAAI,KAAK,OAAO,EAAE,SAAS,EAAE,EAAE,MAAM,GAAG,CAAC,CAAC;AAC9F,YAAM,SAAS,GAAG,YAAY,IAAI,EAAE;AACpC,YAAM,OAAOA,MAAK,aAAa,EAAE;AACjC,UAAI,CAAC,YAAY,OAAO,MAAM,QAAQ,MAAM,OAAO,GAAG,KAAK,QAAQ;AACnE,aAAO,EAAE,MAAM,QAAQ,QAAQ;AAAA,IACjC;AAAA,IAEA,MAAM,SAAS,UAAU,SAAS;AAGhC,YAAM,SAAS,IAAI,CAAC,UAAU,aAAa,GAAG,SAAS,IAAI;AAC3D,UAAI,OAAO,SAAS,GAAG;AACrB,YAAI,CAAC,OAAO,IAAI,GAAG,SAAS,IAAI;AAChC,YAAI,CAAC,UAAU,MAAM,OAAO,GAAG,SAAS,IAAI;AAAA,MAC9C;AACA,aAAO;AAAA,QACL,MAAM;AAAA,QACN,aAAa,SAAS;AAAA,QACtB,SAAS,SAAS;AAAA,QAClB;AAAA,MACF;AAAA,IACF;AAAA,IAEA,MAAM,QAAQ,UAAU;AAGtB,UAAI,CAAC,YAAY,UAAU,WAAW,SAAS,IAAI,GAAG,KAAK,QAAQ;AACnE,UAAI,CAAC,UAAU,MAAM,SAAS,MAAM,GAAG,KAAK,QAAQ;AAAA,IACtD;AAAA,EACF;AACF;AAKO,SAAS,oBAAoB,SAAsB,aAA8B;AACtF,MAAI,WAAW,QAAQ,WAAW,KAAKD,YAAW,QAAQ,WAAW,EAAG,QAAO,QAAQ;AACvF,MAAI,YAAa,QAAOC,MAAK,aAAa,SAAS,QAAQ,WAAW,CAAC;AACvE,SAAO,QAAQ;AACjB;","names":["existsSync","join"]}
+{"version":3,"sources":["../../src/campaign/drivers/skillopt.ts","../../src/campaign/labeled-store/fs-adapter.ts","../../src/campaign/worktree/index.ts"],"sourcesContent":["/**\n * @experimental\n *\n * `skillOptDriver` — a section-aware, bounded-edit `ImprovementDriver` for\n * structured natural-language procedures (SKILL.md files, runbooks, sectioned\n * system prompts, judge rubrics with dimensions). Implements the SkillOpt\n * methodology (Microsoft, 2026): treat the skill document as a trainable\n * optimization target, train the procedure not the weights, constrain each\n * generation to ≤N targeted edits to prevent useful-rule overwrites.\n *\n * Differs from `gepaDriver` in two specific ways:\n *\n * 1. **Bounded edits.** Each candidate must differ from the baseline by at\n *    most `editBudget` sentence-level changes. The \"edit budget functions\n *    as a textual learning rate\" — without it, an LLM proposal can rewrite\n *    so much that useful prior rules get overwritten.\n *\n * 2. **Section preservation.** When the surface is a structured doc, the\n *    H2 headers (and an opt-in `preserveSections` allowlist) are\n *    load-bearing for discoverability. Candidates that delete or rename\n *    preserved sections are rejected at parse time.\n *\n * Selectable alongside `gepaDriver` and `evolutionaryDriver`. Use this when\n * the surface IS a structured doc; use `gepaDriver` when the surface is\n * unstructured prose.\n */\n\nimport { callLlm, type LlmClientOptions } from '../../llm-client'\nimport {\n  buildReflectionPrompt,\n  parseReflectionResponse,\n  type TrialTrace,\n} from '../../reflective-mutation'\nimport type { ImprovementDriver, MutableSurface, ProposeContext } from '../types'\n\nconst REFLECTION_SYSTEM =\n  'You are an expert prompt engineer applying the SkillOpt methodology. ' +\n  'You will edit a structured natural-language procedure under TWO HARD ' +\n  'CONSTRAINTS: (1) preserve every H2 section heading verbatim — do NOT ' +\n  'delete, rename, or merge sections; (2) make at most EDIT_BUDGET targeted ' +\n  'sentence-level edits per candidate — bounded edits prevent overwriting ' +\n  'useful prior rules. Output ONLY a JSON object of shape ' +\n  '{\"proposals\":[{\"label\":string,\"rationale\":string,\"payload\":string}]} ' +\n  'where each `payload` is the FULL improved skill text. No prose outside the JSON.'\n\nexport interface SkillOptDriverOptions {\n  /** Router transport (apiKey/baseUrl). */\n  llm: LlmClientOptions\n  /** Model that performs the reflection. */\n  model: string\n  /** What is being optimized — appears in the reflection prompt for orientation. */\n  target: string\n\n  /** Max edits per generation — SkillOpt's \"textual learning rate\".\n   *  Default 3. Lower = more conservative, higher = more exploratory. */\n  editBudget?: number\n\n  /** Section headings the driver MUST preserve. When the surface is a\n   *  structured skill doc, sections are load-bearing for discoverability.\n   *  Default: auto-detected from H2 headers in the baseline. */\n  preserveSections?: string[]\n\n  /** Surface-specific mutation levers offered to the model. */\n  mutationPrimitives?: string[]\n  /** Top/bottom scenarios surfaced as evidence each generation. Default 3. */\n  evidenceK?: number\n  /** Reflection sampling temperature. Default 0.7. */\n  temperature?: number\n  /** Reflection max tokens. Default 6000. */\n  maxTokens?: number\n}\n\n/** Internal — exported for tests. */\nexport function extractH2Sections(text: string): string[] {\n  const out: string[] = []\n  for (const line of text.split('\\n')) {\n    const match = /^##\\s+(.+?)\\s*$/.exec(line)\n    if (match) out.push(match[1]!)\n  }\n  return out\n}\n\n/** Sentence-level edit distance. Counts distinct sentence add/remove/replace\n *  ops between baseline and candidate using a normalised line-by-line diff.\n *  Imperfect (treats trivial whitespace as identical) but tight enough to\n *  bound an LLM rewrite. Exported for tests. */\nexport function countSentenceEdits(baseline: string, candidate: string): number {\n  const norm = (s: string) =>\n    s\n      .split(/(?<=[.!?])\\s+|\\n/g)\n      .map((p) => p.trim())\n      .filter((p) => p.length > 0)\n  const a = new Set(norm(baseline))\n  const b = new Set(norm(candidate))\n  let edits = 0\n  for (const s of a) if (!b.has(s)) edits++ // deletions\n  for (const s of b) if (!a.has(s)) edits++ // additions\n  return edits\n}\n\nexport function skillOptDriver(opts: SkillOptDriverOptions): ImprovementDriver {\n  const evidenceK = opts.evidenceK ?? 3\n  const editBudget = opts.editBudget ?? 3\n  if (editBudget < 1) {\n    throw new Error(\n      `skillOptDriver: editBudget must be >= 1, got ${editBudget} (use evolutionaryDriver with a noop mutator for measure-only runs)`,\n    )\n  }\n  return {\n    kind: 'skillopt',\n    async propose(ctx: ProposeContext): Promise<MutableSurface[]> {\n      if (typeof ctx.currentSurface !== 'string') {\n        throw new Error(\n          `skillOptDriver: surface must be a string skill document; got ${typeof ctx.currentSurface}. Use evolutionaryDriver with a CodeSurface mutator for code-tier surfaces.`,\n        )\n      }\n      const baseline = ctx.currentSurface\n      const preserveSections = opts.preserveSections ?? extractH2Sections(baseline)\n\n      const { top, bottom, target } = buildEvidence(ctx, evidenceK, opts.target)\n\n      const reflectionUser = buildReflectionPrompt({\n        target,\n        parentPayload: baseline,\n        topTrials: top,\n        bottomTrials: bottom,\n        childCount: ctx.populationSize,\n        mutationPrimitives: opts.mutationPrimitives,\n      })\n      const constraintPreamble = [\n        '',\n        '## SkillOpt constraints (hard rules — violations rejected)',\n        '',\n        `- Edit budget: at most ${editBudget} sentence-level edits per candidate.`,\n        '- Section preservation: every H2 heading below must appear unchanged in your output.',\n        ...preserveSections.map((s) => `  - \\`## ${s}\\``),\n        '',\n        'Reject any candidate in your own thinking that would delete a section, rename a heading, or exceed the edit budget. Make TARGETED, surgical edits — not rewrites.',\n        '',\n      ].join('\\n')\n      const userPrompt = `${reflectionUser}${constraintPreamble}`\n      const system = REFLECTION_SYSTEM.replace('EDIT_BUDGET', String(editBudget))\n\n      const result = await callLlm(\n        {\n          model: opts.model,\n          messages: [\n            { role: 'system', content: system },\n            { role: 'user', content: userPrompt },\n          ],\n          jsonMode: true,\n          temperature: opts.temperature ?? 0.7,\n          maxTokens: opts.maxTokens ?? 6000,\n        },\n        opts.llm,\n      )\n\n      const proposals = parseReflectionResponse(result.content, ctx.populationSize)\n      const out: MutableSurface[] = []\n      for (const proposal of proposals) {\n        const text = typeof proposal.payload === 'string' ? proposal.payload.trim() : ''\n        if (!text || text === baseline) continue\n        if (!validateSections(text, preserveSections)) continue\n        if (countSentenceEdits(baseline, text) > editBudget * 2) continue // x2: add+remove pair per edit\n        if (out.includes(text)) continue\n        out.push(text)\n      }\n      return out\n    },\n  }\n}\n\nfunction validateSections(candidate: string, required: string[]): boolean {\n  if (required.length === 0) return true\n  const have = new Set(extractH2Sections(candidate))\n  for (const section of required) {\n    if (!have.has(section)) return false\n  }\n  return true\n}\n\n/** Reused from gepaDriver pattern — build evidence from prior best candidate. */\nfunction buildEvidence(\n  ctx: ProposeContext,\n  evidenceK: number,\n  baseTarget: string,\n): { top: TrialTrace[]; bottom: TrialTrace[]; target: string } {\n  const last = ctx.history.at(-1)\n  if (!last || last.candidates.length === 0) {\n    return { top: [], bottom: [], target: baseTarget }\n  }\n  const best = [...last.candidates].sort((a, b) => b.composite - a.composite)[0]\n  if (!best) return { top: [], bottom: [], target: baseTarget }\n\n  const byScore = [...best.scenarios].sort((a, b) => b.composite - a.composite)\n  const toTrace = (s: { scenarioId: string; composite: number }): TrialTrace => ({\n    id: s.scenarioId,\n    score: s.composite,\n  })\n  const top = byScore.slice(0, evidenceK).map(toTrace)\n  const bottom = byScore.slice(-evidenceK).reverse().map(toTrace)\n\n  const weakest = Object.entries(best.dimensions)\n    .sort((a, b) => a[1] - b[1])\n    .slice(0, 3)\n    .map(([dim, value]) => `${dim} (${value.toFixed(2)})`)\n  const target =\n    weakest.length > 0 ? `${baseTarget} — weakest dimensions: ${weakest.join(', ')}` : baseTarget\n\n  return { top, bottom, target }\n}\n","/**\n * @experimental\n *\n * Filesystem `LabeledScenarioStore` adapter. The default capture sink for\n * traces + eval artifacts. Production deployments typically swap for a\n * Turso/SQLite adapter (same interface).\n *\n * Records land as one JSONL file per source under `<root>/<source>.jsonl`.\n * Each line is a `LabeledScenarioRecord`. Append-only — no in-place edits.\n *\n * Safety properties enforced at write-time:\n *\n *   - **Provenance required**: writes without `source`, `sourceVersionHash`,\n *     `capturedAt`, `redactionStatus` are rejected. Closes the alignment\n *     reviewer's data-poisoning gap.\n *   - **Per-source rate limits**: optional `rateLimitBucket` + `maxWritesPerMinute`\n *     stops a single tenant/source from flooding the store.\n *\n * Safety properties enforced at sample-time:\n *\n *   - **Required split + capturedBefore**: substrate refuses to sample without\n *     an explicit `split` ('train' | 'test') AND a temporal cutoff. Eliminates\n *     accidental train/test contamination.\n *   - **Default training-source filter**: when the store is sampled with\n *     `split: 'train'`, production-trace records are EXCLUDED unless the\n *     caller passes `filter.source: 'production-trace'` explicitly. Closes\n *     the contamination-by-default gap flagged by the senior eval engineer.\n */\n\nimport { createHash } from 'node:crypto'\nimport { existsSync, mkdirSync, readFileSync, writeFileSync } from 'node:fs'\nimport { join } from 'node:path'\nimport type {\n  LabeledScenarioRecord,\n  LabeledScenarioSampleArgs,\n  LabeledScenarioSource,\n  LabeledScenarioStore,\n  LabeledScenarioWrite,\n} from '../types'\n\nexport interface FsLabeledScenarioStoreOptions {\n  /** Root directory for JSONL files. Created if missing. */\n  root: string\n  /** Per-source rate limit. When set, writes exceeding the cap are rejected\n   *  with a typed error. Default: no limit. */\n  maxWritesPerMinutePerBucket?: number\n  /** Test seam — override `Date.now()` for deterministic tests. */\n  now?: () => number\n}\n\nexport class LabeledScenarioStoreError extends Error {\n  constructor(\n    public readonly code: string,\n    message: string,\n  ) {\n    super(message)\n    this.name = 'LabeledScenarioStoreError'\n  }\n}\n\ninterface RateLimitState {\n  bucket: string\n  windowStartMs: number\n  count: number\n}\n\nexport class FsLabeledScenarioStore implements LabeledScenarioStore {\n  private readonly now: () => number\n  private readonly rateLimits = new Map<string, RateLimitState>()\n\n  constructor(private readonly options: FsLabeledScenarioStoreOptions) {\n    if (!existsSync(options.root)) mkdirSync(options.root, { recursive: true })\n    this.now = options.now ?? Date.now\n  }\n\n  async observe(write: LabeledScenarioWrite): Promise<void> {\n    this.assertProvenance(write)\n    this.assertRateLimit(write)\n    const record = this.toRecord(write)\n    const path = this.pathForSource(write.source)\n    const line = `${JSON.stringify(record)}\\n`\n    // Append atomically. For high-throughput a writev-friendly buffered\n    // implementation lands in the Turso adapter; FS adapter is for tests +\n    // local dev + small workloads.\n    appendLine(path, line)\n  }\n\n  async sample(args: LabeledScenarioSampleArgs): Promise<LabeledScenarioRecord[]> {\n    if (!args.split) {\n      throw new LabeledScenarioStoreError(\n        'split_required',\n        'sample() requires an explicit `split` (train | test) — substrate refuses ambiguous reads',\n      )\n    }\n    if (!args.capturedBefore) {\n      throw new LabeledScenarioStoreError(\n        'capturedBefore_required',\n        'sample() requires an explicit `capturedBefore` timestamp for temporal-split discipline',\n      )\n    }\n\n    const all: LabeledScenarioRecord[] = []\n    for (const source of ALL_SOURCES) {\n      // Default training-source filter: when sampling train, EXCLUDE\n      // production-trace records unless the caller asks for them.\n      if (args.split === 'train' && source === 'production-trace') {\n        const explicit = sourceFilterContains(args.filter?.source, 'production-trace')\n        if (!explicit) continue\n      }\n      const path = this.pathForSource(source)\n      if (!existsSync(path)) continue\n      const lines = readFileSync(path, 'utf8').split('\\n').filter(Boolean)\n      for (const line of lines) {\n        let record: LabeledScenarioRecord\n        try {\n          record = JSON.parse(line) as LabeledScenarioRecord\n        } catch {\n          continue\n        }\n        if (!matchesFilter(record, args, source)) continue\n        all.push(record)\n      }\n    }\n\n    // Deterministic order: by capturedAt ascending, then recordHash.\n    all.sort((a, b) => {\n      if (a.capturedAt !== b.capturedAt) return a.capturedAt.localeCompare(b.capturedAt)\n      return a.recordHash.localeCompare(b.recordHash)\n    })\n\n    return all.slice(0, args.count)\n  }\n\n  async size(): Promise<{ train: number; test: number; bySource: Record<string, number> }> {\n    const bySource: Record<string, number> = {}\n    let total = 0\n    for (const source of ALL_SOURCES) {\n      const path = this.pathForSource(source)\n      if (!existsSync(path)) {\n        bySource[source] = 0\n        continue\n      }\n      const count = readFileSync(path, 'utf8').split('\\n').filter(Boolean).length\n      bySource[source] = count\n      total += count\n    }\n    // FS adapter doesn't track split assignments per-record (split is\n    // computed at sample-time based on `capturedBefore`). For size(), we\n    // report `train`+`test` as the same total — split is a sampling concept.\n    return { train: total, test: total, bySource }\n  }\n\n  private assertProvenance(write: LabeledScenarioWrite): void {\n    if (!write.source) {\n      throw new LabeledScenarioStoreError(\n        'missing_source',\n        'LabeledScenarioWrite requires `source`',\n      )\n    }\n    if (!write.sourceVersionHash || write.sourceVersionHash.length === 0) {\n      throw new LabeledScenarioStoreError(\n        'missing_source_version',\n        'LabeledScenarioWrite requires `sourceVersionHash` (git sha or substrate version)',\n      )\n    }\n    if (!write.capturedAt) {\n      throw new LabeledScenarioStoreError(\n        'missing_captured_at',\n        'LabeledScenarioWrite requires `capturedAt` ISO timestamp',\n      )\n    }\n    if (!write.redactionStatus) {\n      throw new LabeledScenarioStoreError(\n        'missing_redaction_status',\n        'LabeledScenarioWrite requires explicit `redactionStatus` — raw / redacted-pii / redacted-secrets / fully-redacted',\n      )\n    }\n    if (!ALL_SOURCES.includes(write.source)) {\n      throw new LabeledScenarioStoreError(\n        'unknown_source',\n        `LabeledScenarioWrite.source must be one of: ${ALL_SOURCES.join(', ')}`,\n      )\n    }\n  }\n\n  private assertRateLimit(write: LabeledScenarioWrite): void {\n    const cap = this.options.maxWritesPerMinutePerBucket\n    if (!cap || !write.rateLimitBucket) return\n    const now = this.now()\n    const windowMs = 60_000\n    let state = this.rateLimits.get(write.rateLimitBucket)\n    if (!state || now - state.windowStartMs >= windowMs) {\n      state = { bucket: write.rateLimitBucket, windowStartMs: now, count: 0 }\n      this.rateLimits.set(write.rateLimitBucket, state)\n    }\n    if (state.count >= cap) {\n      throw new LabeledScenarioStoreError(\n        'rate_limit_exceeded',\n        `LabeledScenarioStore: bucket ${write.rateLimitBucket} exceeded ${cap} writes/min`,\n      )\n    }\n    state.count += 1\n  }\n\n  private toRecord(write: LabeledScenarioWrite): LabeledScenarioRecord {\n    const recordHash = sha256(\n      JSON.stringify({\n        id: write.scenario.id,\n        src: write.source,\n        at: write.capturedAt,\n        ver: write.sourceVersionHash,\n      }),\n    )\n    // FS adapter assigns split at sample-time, but we cache a hint here\n    // based on capturedAt vs the world's \"now\" — sampler overrides this.\n    return {\n      ...write,\n      recordHash,\n      split: 'train',\n    }\n  }\n\n  private pathForSource(source: string): string {\n    return join(this.options.root, `${source}.jsonl`)\n  }\n}\n\nconst ALL_SOURCES: LabeledScenarioWrite['source'][] = [\n  'production-trace',\n  'eval-run',\n  'manual',\n  'red-team',\n  'synthetic',\n]\n\nfunction sourceFilterContains(\n  filter: LabeledScenarioSource | LabeledScenarioSource[] | undefined,\n  needle: LabeledScenarioSource,\n): boolean {\n  if (!filter) return false\n  if (Array.isArray(filter)) return filter.includes(needle)\n  return filter === needle\n}\n\nfunction matchesFilter(\n  record: LabeledScenarioRecord,\n  args: LabeledScenarioSampleArgs,\n  source: string,\n): boolean {\n  // Temporal cutoff — train must be capturedAt < capturedBefore.\n  if (args.split === 'train' && record.capturedAt >= args.capturedBefore) return false\n  if (args.split === 'test' && record.capturedAt < args.capturedBefore) return false\n\n  const f = args.filter\n  if (!f) return true\n  if (f.kind && record.scenario.kind !== f.kind) return false\n  if (f.source) {\n    const sources = Array.isArray(f.source) ? f.source : [f.source]\n    if (!sources.includes(source as never)) return false\n  }\n  if (f.minComposite !== undefined || f.maxComposite !== undefined) {\n    const composites = Object.values(record.judgeScores).map((s) => s.composite)\n    const max = composites.length === 0 ? 0 : Math.max(...composites)\n    if (f.minComposite !== undefined && max < f.minComposite) return false\n    if (f.maxComposite !== undefined && max > f.maxComposite) return false\n  }\n  return true\n}\n\nfunction sha256(input: string): string {\n  return createHash('sha256').update(input).digest('hex').slice(0, 16)\n}\n\nfunction appendLine(path: string, line: string): void {\n  if (existsSync(path)) {\n    const existing = readFileSync(path, 'utf8')\n    writeFileSync(path, existing + line)\n  } else {\n    writeFileSync(path, line)\n  }\n}\n","/**\n * @experimental\n *\n * VCS-pluggable worktree adapter. One improvement = one worktree, PR-like\n * (multiple commits allowed). A code-tier driver's `propose()` creates a\n * worktree, an agent commits the change into it, and `finalize()` returns a\n * `CodeSurface{ worktreeRef }` the measurement checks out to run the worker\n * against the changed code. On promotion the worktree becomes the PR branch.\n *\n * The interface is VCS-agnostic so a future `jj` ([jj-vcs](https://github.com/jj-vcs/jj))\n * adapter can slot in without touching driver code. Only the git adapter\n * ships today. See `docs/design/self-improvement-engine.md`.\n */\n\nimport { execFileSync } from 'node:child_process'\nimport { existsSync } from 'node:fs'\nimport { basename, isAbsolute, join } from 'node:path'\nimport type { CodeSurface } from '../types'\n\nexport interface Worktree {\n  /** Absolute path to the checked-out worktree directory. */\n  path: string\n  /** The branch the worktree is on (becomes the PR branch on promotion). */\n  branch: string\n  /** The ref the worktree was forked from. */\n  baseRef: string\n}\n\nexport interface WorktreeAdapter {\n  /** Create an isolated worktree on a fresh branch off `baseRef`. */\n  create(opts: { baseRef: string; label: string }): Promise<Worktree>\n  /** Commit any pending changes in the worktree, then return a CodeSurface\n   *  pointing at it. The agent has already written its change into\n   *  `worktree.path` by the time this is called. */\n  finalize(worktree: Worktree, summary: string): Promise<CodeSurface>\n  /** Remove the worktree (and its branch) — called for losing candidates. */\n  discard(worktree: Worktree): Promise<void>\n}\n\nexport class WorktreeAdapterError extends Error {\n  constructor(\n    message: string,\n    readonly cause?: unknown,\n  ) {\n    super(message)\n    this.name = 'WorktreeAdapterError'\n  }\n}\n\nexport interface GitWorktreeAdapterOptions {\n  /** Repo root the worktrees fork from. */\n  repoRoot: string\n  /** Directory worktrees are created under. Default: `<repoRoot>/.worktrees`. */\n  worktreeDir?: string\n  /** Branch-name prefix. Default: `improve`. */\n  branchPrefix?: string\n  /** Test seam — defaults to a real `git` runner. */\n  git?: (args: string[], cwd: string) => string\n}\n\nfunction defaultGit(args: string[], cwd: string): string {\n  try {\n    return execFileSync('git', args, { cwd, encoding: 'utf8' }).trim()\n  } catch (err) {\n    const stderr =\n      err && typeof err === 'object' && 'stderr' in err\n        ? String((err as { stderr: unknown }).stderr)\n        : ''\n    throw new WorktreeAdapterError(`git ${args.join(' ')} failed: ${stderr || String(err)}`, err)\n  }\n}\n\n/** Slugify a label into a branch-safe segment. */\nfunction slug(label: string): string {\n  return (\n    label\n      .toLowerCase()\n      .replace(/[^a-z0-9]+/g, '-')\n      .replace(/^-+|-+$/g, '')\n      .slice(0, 48) || 'candidate'\n  )\n}\n\nexport function gitWorktreeAdapter(opts: GitWorktreeAdapterOptions): WorktreeAdapter {\n  const git = opts.git ?? defaultGit\n  const worktreeDir = opts.worktreeDir ?? join(opts.repoRoot, '.worktrees')\n  const branchPrefix = opts.branchPrefix ?? 'improve'\n\n  return {\n    async create({ baseRef, label }) {\n      const id = `${slug(label)}-${Date.now().toString(36)}-${Math.random().toString(36).slice(2, 6)}`\n      const branch = `${branchPrefix}/${id}`\n      const path = join(worktreeDir, id)\n      git(['worktree', 'add', '-b', branch, path, baseRef], opts.repoRoot)\n      return { path, branch, baseRef }\n    },\n\n    async finalize(worktree, summary) {\n      // Stage + commit any pending changes the agent left in the worktree.\n      // A no-op commit is refused by git, so only commit when the tree is dirty.\n      const status = git(['status', '--porcelain'], worktree.path)\n      if (status.length > 0) {\n        git(['add', '-A'], worktree.path)\n        git(['commit', '-m', summary], worktree.path)\n      }\n      return {\n        kind: 'code',\n        worktreeRef: worktree.path,\n        baseRef: worktree.baseRef,\n        summary,\n      }\n    },\n\n    async discard(worktree) {\n      // Remove the worktree, then delete its branch. Force-remove because the\n      // worktree may hold uncommitted experiment state we're discarding.\n      git(['worktree', 'remove', '--force', worktree.path], opts.repoRoot)\n      git(['branch', '-D', worktree.branch], opts.repoRoot)\n    },\n  }\n}\n\n/** Resolve a `CodeSurface`'s worktreeRef to a directory the measurement can\n *  run the worker in. A path ref is returned as-is; anything else is treated\n *  as a ref under the adapter's worktree dir. */\nexport function resolveWorktreePath(surface: CodeSurface, worktreeDir?: string): string {\n  if (isAbsolute(surface.worktreeRef) && existsSync(surface.worktreeRef)) return surface.worktreeRef\n  if (worktreeDir) return join(worktreeDir, basename(surface.worktreeRef))\n  return surface.worktreeRef\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAmCA,IAAM,oBACJ;AAqCK,SAAS,kBAAkB,MAAwB;AACxD,QAAM,MAAgB,CAAC;AACvB,aAAW,QAAQ,KAAK,MAAM,IAAI,GAAG;AACnC,UAAM,QAAQ,kBAAkB,KAAK,IAAI;AACzC,QAAI,MAAO,KAAI,KAAK,MAAM,CAAC,CAAE;AAAA,EAC/B;AACA,SAAO;AACT;AAMO,SAAS,mBAAmB,UAAkB,WAA2B;AAC9E,QAAM,OAAO,CAAC,MACZ,EACG,MAAM,mBAAmB,EACzB,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC,EACnB,OAAO,CAAC,MAAM,EAAE,SAAS,CAAC;AAC/B,QAAM,IAAI,IAAI,IAAI,KAAK,QAAQ,CAAC;AAChC,QAAM,IAAI,IAAI,IAAI,KAAK,SAAS,CAAC;AACjC,MAAI,QAAQ;AACZ,aAAW,KAAK,EAAG,KAAI,CAAC,EAAE,IAAI,CAAC,EAAG;AAClC,aAAW,KAAK,EAAG,KAAI,CAAC,EAAE,IAAI,CAAC,EAAG;AAClC,SAAO;AACT;AAEO,SAAS,eAAe,MAAgD;AAC7E,QAAM,YAAY,KAAK,aAAa;AACpC,QAAM,aAAa,KAAK,cAAc;AACtC,MAAI,aAAa,GAAG;AAClB,UAAM,IAAI;AAAA,MACR,gDAAgD,UAAU;AAAA,IAC5D;AAAA,EACF;AACA,SAAO;AAAA,IACL,MAAM;AAAA,IACN,MAAM,QAAQ,KAAgD;AAC5D,UAAI,OAAO,IAAI,mBAAmB,UAAU;AAC1C,cAAM,IAAI;AAAA,UACR,gEAAgE,OAAO,IAAI,cAAc;AAAA,QAC3F;AAAA,MACF;AACA,YAAM,WAAW,IAAI;AACrB,YAAM,mBAAmB,KAAK,oBAAoB,kBAAkB,QAAQ;AAE5E,YAAM,EAAE,KAAK,QAAQ,OAAO,IAAI,cAAc,KAAK,WAAW,KAAK,MAAM;AAEzE,YAAM,iBAAiB,sBAAsB;AAAA,QAC3C;AAAA,QACA,eAAe;AAAA,QACf,WAAW;AAAA,QACX,cAAc;AAAA,QACd,YAAY,IAAI;AAAA,QAChB,oBAAoB,KAAK;AAAA,MAC3B,CAAC;AACD,YAAM,qBAAqB;AAAA,QACzB;AAAA,QACA;AAAA,QACA;AAAA,QACA,0BAA0B,UAAU;AAAA,QACpC;AAAA,QACA,GAAG,iBAAiB,IAAI,CAAC,MAAM,YAAY,CAAC,IAAI;AAAA,QAChD;AAAA,QACA;AAAA,QACA;AAAA,MACF,EAAE,KAAK,IAAI;AACX,YAAM,aAAa,GAAG,cAAc,GAAG,kBAAkB;AACzD,YAAM,SAAS,kBAAkB,QAAQ,eAAe,OAAO,UAAU,CAAC;AAE1E,YAAM,SAAS,MAAM;AAAA,QACnB;AAAA,UACE,OAAO,KAAK;AAAA,UACZ,UAAU;AAAA,YACR,EAAE,MAAM,UAAU,SAAS,OAAO;AAAA,YAClC,EAAE,MAAM,QAAQ,SAAS,WAAW;AAAA,UACtC;AAAA,UACA,UAAU;AAAA,UACV,aAAa,KAAK,eAAe;AAAA,UACjC,WAAW,KAAK,aAAa;AAAA,QAC/B;AAAA,QACA,KAAK;AAAA,MACP;AAEA,YAAM,YAAY,wBAAwB,OAAO,SAAS,IAAI,cAAc;AAC5E,YAAM,MAAwB,CAAC;AAC/B,iBAAW,YAAY,WAAW;AAChC,cAAM,OAAO,OAAO,SAAS,YAAY,WAAW,SAAS,QAAQ,KAAK,IAAI;AAC9E,YAAI,CAAC,QAAQ,SAAS,SAAU;AAChC,YAAI,CAAC,iBAAiB,MAAM,gBAAgB,EAAG;AAC/C,YAAI,mBAAmB,UAAU,IAAI,IAAI,aAAa,EAAG;AACzD,YAAI,IAAI,SAAS,IAAI,EAAG;AACxB,YAAI,KAAK,IAAI;AAAA,MACf;AACA,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAEA,SAAS,iBAAiB,WAAmB,UAA6B;AACxE,MAAI,SAAS,WAAW,EAAG,QAAO;AAClC,QAAM,OAAO,IAAI,IAAI,kBAAkB,SAAS,CAAC;AACjD,aAAW,WAAW,UAAU;AAC9B,QAAI,CAAC,KAAK,IAAI,OAAO,EAAG,QAAO;AAAA,EACjC;AACA,SAAO;AACT;AAGA,SAAS,cACP,KACA,WACA,YAC6D;AAC7D,QAAM,OAAO,IAAI,QAAQ,GAAG,EAAE;AAC9B,MAAI,CAAC,QAAQ,KAAK,WAAW,WAAW,GAAG;AACzC,WAAO,EAAE,KAAK,CAAC,GAAG,QAAQ,CAAC,GAAG,QAAQ,WAAW;AAAA,EACnD;AACA,QAAM,OAAO,CAAC,GAAG,KAAK,UAAU,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS,EAAE,CAAC;AAC7E,MAAI,CAAC,KAAM,QAAO,EAAE,KAAK,CAAC,GAAG,QAAQ,CAAC,GAAG,QAAQ,WAAW;AAE5D,QAAM,UAAU,CAAC,GAAG,KAAK,SAAS,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAC5E,QAAM,UAAU,CAAC,OAA8D;AAAA,IAC7E,IAAI,EAAE;AAAA,IACN,OAAO,EAAE;AAAA,EACX;AACA,QAAM,MAAM,QAAQ,MAAM,GAAG,SAAS,EAAE,IAAI,OAAO;AACnD,QAAM,SAAS,QAAQ,MAAM,CAAC,SAAS,EAAE,QAAQ,EAAE,IAAI,OAAO;AAE9D,QAAM,UAAU,OAAO,QAAQ,KAAK,UAAU,EAC3C,KAAK,CAAC,GAAG,MAAM,EAAE,CAAC,IAAI,EAAE,CAAC,CAAC,EAC1B,MAAM,GAAG,CAAC,EACV,IAAI,CAAC,CAAC,KAAK,KAAK,MAAM,GAAG,GAAG,KAAK,MAAM,QAAQ,CAAC,CAAC,GAAG;AACvD,QAAM,SACJ,QAAQ,SAAS,IAAI,GAAG,UAAU,+BAA0B,QAAQ,KAAK,IAAI,CAAC,KAAK;AAErF,SAAO,EAAE,KAAK,QAAQ,OAAO;AAC/B;;;ACrLA,SAAS,kBAAkB;AAC3B,SAAS,YAAY,WAAW,cAAc,qBAAqB;AACnE,SAAS,YAAY;AAmBd,IAAM,4BAAN,cAAwC,MAAM;AAAA,EACnD,YACkB,MAChB,SACA;AACA,UAAM,OAAO;AAHG;AAIhB,SAAK,OAAO;AAAA,EACd;AAAA,EALkB;AAMpB;AAQO,IAAM,yBAAN,MAA6D;AAAA,EAIlE,YAA6B,SAAwC;AAAxC;AAC3B,QAAI,CAAC,WAAW,QAAQ,IAAI,EAAG,WAAU,QAAQ,MAAM,EAAE,WAAW,KAAK,CAAC;AAC1E,SAAK,MAAM,QAAQ,OAAO,KAAK;AAAA,EACjC;AAAA,EAH6B;AAAA,EAHZ;AAAA,EACA,aAAa,oBAAI,IAA4B;AAAA,EAO9D,MAAM,QAAQ,OAA4C;AACxD,SAAK,iBAAiB,KAAK;AAC3B,SAAK,gBAAgB,KAAK;AAC1B,UAAM,SAAS,KAAK,SAAS,KAAK;AAClC,UAAM,OAAO,KAAK,cAAc,MAAM,MAAM;AAC5C,UAAM,OAAO,GAAG,KAAK,UAAU,MAAM,CAAC;AAAA;AAItC,eAAW,MAAM,IAAI;AAAA,EACvB;AAAA,EAEA,MAAM,OAAO,MAAmE;AAC9E,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,KAAK,gBAAgB;AACxB,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AAEA,UAAM,MAA+B,CAAC;AACtC,eAAW,UAAU,aAAa;AAGhC,UAAI,KAAK,UAAU,WAAW,WAAW,oBAAoB;AAC3D,cAAM,WAAW,qBAAqB,KAAK,QAAQ,QAAQ,kBAAkB;AAC7E,YAAI,CAAC,SAAU;AAAA,MACjB;AACA,YAAM,OAAO,KAAK,cAAc,MAAM;AACtC,UAAI,CAAC,WAAW,IAAI,EAAG;AACvB,YAAM,QAAQ,aAAa,MAAM,MAAM,EAAE,MAAM,IAAI,EAAE,OAAO,OAAO;AACnE,iBAAW,QAAQ,OAAO;AACxB,YAAI;AACJ,YAAI;AACF,mBAAS,KAAK,MAAM,IAAI;AAAA,QAC1B,QAAQ;AACN;AAAA,QACF;AACA,YAAI,CAAC,cAAc,QAAQ,MAAM,MAAM,EAAG;AAC1C,YAAI,KAAK,MAAM;AAAA,MACjB;AAAA,IACF;AAGA,QAAI,KAAK,CAAC,GAAG,MAAM;AACjB,UAAI,EAAE,eAAe,EAAE,WAAY,QAAO,EAAE,WAAW,cAAc,EAAE,UAAU;AACjF,aAAO,EAAE,WAAW,cAAc,EAAE,UAAU;AAAA,IAChD,CAAC;AAED,WAAO,IAAI,MAAM,GAAG,KAAK,KAAK;AAAA,EAChC;AAAA,EAEA,MAAM,OAAmF;AACvF,UAAM,WAAmC,CAAC;AAC1C,QAAI,QAAQ;AACZ,eAAW,UAAU,aAAa;AAChC,YAAM,OAAO,KAAK,cAAc,MAAM;AACtC,UAAI,CAAC,WAAW,IAAI,GAAG;AACrB,iBAAS,MAAM,IAAI;AACnB;AAAA,MACF;AACA,YAAM,QAAQ,aAAa,MAAM,MAAM,EAAE,MAAM,IAAI,EAAE,OAAO,OAAO,EAAE;AACrE,eAAS,MAAM,IAAI;AACnB,eAAS;AAAA,IACX;AAIA,WAAO,EAAE,OAAO,OAAO,MAAM,OAAO,SAAS;AAAA,EAC/C;AAAA,EAEQ,iBAAiB,OAAmC;AAC1D,QAAI,CAAC,MAAM,QAAQ;AACjB,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,MAAM,qBAAqB,MAAM,kBAAkB,WAAW,GAAG;AACpE,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,MAAM,YAAY;AACrB,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,MAAM,iBAAiB;AAC1B,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,YAAY,SAAS,MAAM,MAAM,GAAG;AACvC,YAAM,IAAI;AAAA,QACR;AAAA,QACA,+CAA+C,YAAY,KAAK,IAAI,CAAC;AAAA,MACvE;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,gBAAgB,OAAmC;AACzD,UAAM,MAAM,KAAK,QAAQ;AACzB,QAAI,CAAC,OAAO,CAAC,MAAM,gBAAiB;AACpC,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,WAAW;AACjB,QAAI,QAAQ,KAAK,WAAW,IAAI,MAAM,eAAe;AACrD,QAAI,CAAC,SAAS,MAAM,MAAM,iBAAiB,UAAU;AACnD,cAAQ,EAAE,QAAQ,MAAM,iBAAiB,eAAe,KAAK,OAAO,EAAE;AACtE,WAAK,WAAW,IAAI,MAAM,iBAAiB,KAAK;AAAA,IAClD;AACA,QAAI,MAAM,SAAS,KAAK;AACtB,YAAM,IAAI;AAAA,QACR;AAAA,QACA,gCAAgC,MAAM,eAAe,aAAa,GAAG;AAAA,MACvE;AAAA,IACF;AACA,UAAM,SAAS;AAAA,EACjB;AAAA,EAEQ,SAAS,OAAoD;AACnE,UAAM,aAAa;AAAA,MACjB,KAAK,UAAU;AAAA,QACb,IAAI,MAAM,SAAS;AAAA,QACnB,KAAK,MAAM;AAAA,QACX,IAAI,MAAM;AAAA,QACV,KAAK,MAAM;AAAA,MACb,CAAC;AAAA,IACH;AAGA,WAAO;AAAA,MACL,GAAG;AAAA,MACH;AAAA,MACA,OAAO;AAAA,IACT;AAAA,EACF;AAAA,EAEQ,cAAc,QAAwB;AAC5C,WAAO,KAAK,KAAK,QAAQ,MAAM,GAAG,MAAM,QAAQ;AAAA,EAClD;AACF;AAEA,IAAM,cAAgD;AAAA,EACpD;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAEA,SAAS,qBACP,QACA,QACS;AACT,MAAI,CAAC,OAAQ,QAAO;AACpB,MAAI,MAAM,QAAQ,MAAM,EAAG,QAAO,OAAO,SAAS,MAAM;AACxD,SAAO,WAAW;AACpB;AAEA,SAAS,cACP,QACA,MACA,QACS;AAET,MAAI,KAAK,UAAU,WAAW,OAAO,cAAc,KAAK,eAAgB,QAAO;AAC/E,MAAI,KAAK,UAAU,UAAU,OAAO,aAAa,KAAK,eAAgB,QAAO;AAE7E,QAAM,IAAI,KAAK;AACf,MAAI,CAAC,EAAG,QAAO;AACf,MAAI,EAAE,QAAQ,OAAO,SAAS,SAAS,EAAE,KAAM,QAAO;AACtD,MAAI,EAAE,QAAQ;AACZ,UAAM,UAAU,MAAM,QAAQ,EAAE,MAAM,IAAI,EAAE,SAAS,CAAC,EAAE,MAAM;AAC9D,QAAI,CAAC,QAAQ,SAAS,MAAe,EAAG,QAAO;AAAA,EACjD;AACA,MAAI,EAAE,iBAAiB,UAAa,EAAE,iBAAiB,QAAW;AAChE,UAAM,aAAa,OAAO,OAAO,OAAO,WAAW,EAAE,IAAI,CAAC,MAAM,EAAE,SAAS;AAC3E,UAAM,MAAM,WAAW,WAAW,IAAI,IAAI,KAAK,IAAI,GAAG,UAAU;AAChE,QAAI,EAAE,iBAAiB,UAAa,MAAM,EAAE,aAAc,QAAO;AACjE,QAAI,EAAE,iBAAiB,UAAa,MAAM,EAAE,aAAc,QAAO;AAAA,EACnE;AACA,SAAO;AACT;AAEA,SAAS,OAAO,OAAuB;AACrC,SAAO,WAAW,QAAQ,EAAE,OAAO,KAAK,EAAE,OAAO,KAAK,EAAE,MAAM,GAAG,EAAE;AACrE;AAEA,SAAS,WAAW,MAAc,MAAoB;AACpD,MAAI,WAAW,IAAI,GAAG;AACpB,UAAM,WAAW,aAAa,MAAM,MAAM;AAC1C,kBAAc,MAAM,WAAW,IAAI;AAAA,EACrC,OAAO;AACL,kBAAc,MAAM,IAAI;AAAA,EAC1B;AACF;;;AC1QA,SAAS,oBAAoB;AAC7B,SAAS,cAAAA,mBAAkB;AAC3B,SAAS,UAAU,YAAY,QAAAC,aAAY;AAuBpC,IAAM,uBAAN,cAAmC,MAAM;AAAA,EAC9C,YACE,SACS,OACT;AACA,UAAM,OAAO;AAFJ;AAGT,SAAK,OAAO;AAAA,EACd;AAAA,EAJW;AAKb;AAaA,SAAS,WAAW,MAAgB,KAAqB;AACvD,MAAI;AACF,WAAO,aAAa,OAAO,MAAM,EAAE,KAAK,UAAU,OAAO,CAAC,EAAE,KAAK;AAAA,EACnE,SAAS,KAAK;AACZ,UAAM,SACJ,OAAO,OAAO,QAAQ,YAAY,YAAY,MAC1C,OAAQ,IAA4B,MAAM,IAC1C;AACN,UAAM,IAAI,qBAAqB,OAAO,KAAK,KAAK,GAAG,CAAC,YAAY,UAAU,OAAO,GAAG,CAAC,IAAI,GAAG;AAAA,EAC9F;AACF;AAGA,SAAS,KAAK,OAAuB;AACnC,SACE,MACG,YAAY,EACZ,QAAQ,eAAe,GAAG,EAC1B,QAAQ,YAAY,EAAE,EACtB,MAAM,GAAG,EAAE,KAAK;AAEvB;AAEO,SAAS,mBAAmB,MAAkD;AACnF,QAAM,MAAM,KAAK,OAAO;AACxB,QAAM,cAAc,KAAK,eAAeA,MAAK,KAAK,UAAU,YAAY;AACxE,QAAM,eAAe,KAAK,gBAAgB;AAE1C,SAAO;AAAA,IACL,MAAM,OAAO,EAAE,SAAS,MAAM,GAAG;AAC/B,YAAM,KAAK,GAAG,KAAK,KAAK,CAAC,IAAI,KAAK,IAAI,EAAE,SAAS,EAAE,CAAC,IAAI,KAAK,OAAO,EAAE,SAAS,EAAE,EAAE,MAAM,GAAG,CAAC,CAAC;AAC9F,YAAM,SAAS,GAAG,YAAY,IAAI,EAAE;AACpC,YAAM,OAAOA,MAAK,aAAa,EAAE;AACjC,UAAI,CAAC,YAAY,OAAO,MAAM,QAAQ,MAAM,OAAO,GAAG,KAAK,QAAQ;AACnE,aAAO,EAAE,MAAM,QAAQ,QAAQ;AAAA,IACjC;AAAA,IAEA,MAAM,SAAS,UAAU,SAAS;AAGhC,YAAM,SAAS,IAAI,CAAC,UAAU,aAAa,GAAG,SAAS,IAAI;AAC3D,UAAI,OAAO,SAAS,GAAG;AACrB,YAAI,CAAC,OAAO,IAAI,GAAG,SAAS,IAAI;AAChC,YAAI,CAAC,UAAU,MAAM,OAAO,GAAG,SAAS,IAAI;AAAA,MAC9C;AACA,aAAO;AAAA,QACL,MAAM;AAAA,QACN,aAAa,SAAS;AAAA,QACtB,SAAS,SAAS;AAAA,QAClB;AAAA,MACF;AAAA,IACF;AAAA,IAEA,MAAM,QAAQ,UAAU;AAGtB,UAAI,CAAC,YAAY,UAAU,WAAW,SAAS,IAAI,GAAG,KAAK,QAAQ;AACnE,UAAI,CAAC,UAAU,MAAM,SAAS,MAAM,GAAG,KAAK,QAAQ;AAAA,IACtD;AAAA,EACF;AACF;AAKO,SAAS,oBAAoB,SAAsB,aAA8B;AACtF,MAAI,WAAW,QAAQ,WAAW,KAAKD,YAAW,QAAQ,WAAW,EAAG,QAAO,QAAQ;AACvF,MAAI,YAAa,QAAOC,MAAK,aAAa,SAAS,QAAQ,WAAW,CAAC;AACvE,SAAO,QAAQ;AACjB;","names":["existsSync","join"]}

package/dist/openapi.json

@@ -2,7 +2,7 @@
   "openapi": "3.1.0",
   "info": {
     "title": "@tangle-network/agent-eval — wire protocol",
-    "version": "0.50.2",
+    "version": "0.51.0",
     "description": "HTTP and stdio RPC interface to agent-eval. The TypeScript runtime is the source of truth; this spec is the contract that cross-language clients (Python, Rust, Go) generate from.\n\nWire-protocol version: 1.0.0. Bumps on breaking changes to request/response schemas.",
     "contact": {
       "name": "Tangle Network",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/agent-eval",
-  "version": "0.50.2",
+  "version": "0.51.0",
   "description": "Substrate for self-improving agents: traces, verifiable rewards, preferences, GEPA / reflective mutation, auto-research, replay, sequential anytime-valid stats, and release gates.",
   "homepage": "https://github.com/tangle-network/agent-eval#readme",
   "repository": {