@dailephd/my-dev-kit-lab 0.6.2 → 0.7.0

package/benchmarks/projects/context-window-scaling-fixed-ts/src/filler/tier-4/workshop-bench-catalog-43.ts

@@ -0,0 +1,302 @@
+// Scale-only filler (tier-4). Ordinary catalog data for a workshop bench listing.
+// It is unrelated to every task under src/tasks and exists only to add repository context.
+
+export interface WorkshopBenchRecord43 {
+  id: string;
+  label: string;
+  capacity: number;
+  priority: number;
+  active: boolean;
+  notes: string;
+}
+
+export const workshopBenchRecords43: readonly WorkshopBenchRecord43[] = [
+  {
+    id: "wor-43-001",
+    label: "central annex 1",
+    capacity: 74,
+    priority: 5,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-002",
+    label: "northern row 2",
+    capacity: 44,
+    priority: 4,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-003",
+    label: "central row 3",
+    capacity: 78,
+    priority: 1,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-004",
+    label: "central annex 4",
+    capacity: 20,
+    priority: 3,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 5"
+  },
+  {
+    id: "wor-43-005",
+    label: "granite annex 5",
+    capacity: 72,
+    priority: 4,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-006",
+    label: "eastern annex 6",
+    capacity: 58,
+    priority: 3,
+    active: false,
+    notes: "central workshop bench entry reviewed in cycle 9"
+  },
+  {
+    id: "wor-43-007",
+    label: "granite hall 7",
+    capacity: 66,
+    priority: 1,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-008",
+    label: "central row 8",
+    capacity: 44,
+    priority: 2,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 9"
+  },
+  {
+    id: "wor-43-009",
+    label: "northern row 9",
+    capacity: 20,
+    priority: 5,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 5"
+  },
+  {
+    id: "wor-43-010",
+    label: "northern hall 10",
+    capacity: 10,
+    priority: 4,
+    active: true,
+    notes: "northern workshop bench entry reviewed in cycle 5"
+  },
+  {
+    id: "wor-43-011",
+    label: "northern hall 11",
+    capacity: 44,
+    priority: 4,
+    active: false,
+    notes: "central workshop bench entry reviewed in cycle 5"
+  },
+  {
+    id: "wor-43-012",
+    label: "central row 12",
+    capacity: 28,
+    priority: 2,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-013",
+    label: "granite annex 13",
+    capacity: 56,
+    priority: 1,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 5"
+  },
+  {
+    id: "wor-43-014",
+    label: "northern row 14",
+    capacity: 90,
+    priority: 2,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 9"
+  },
+  {
+    id: "wor-43-015",
+    label: "granite hall 15",
+    capacity: 28,
+    priority: 2,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 9"
+  },
+  {
+    id: "wor-43-016",
+    label: "northern hall 16",
+    capacity: 38,
+    priority: 3,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 9"
+  },
+  {
+    id: "wor-43-017",
+    label: "granite annex 17",
+    capacity: 16,
+    priority: 1,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 5"
+  },
+  {
+    id: "wor-43-018",
+    label: "northern annex 18",
+    capacity: 52,
+    priority: 2,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 9"
+  },
+  {
+    id: "wor-43-019",
+    label: "central row 19",
+    capacity: 32,
+    priority: 5,
+    active: false,
+    notes: "central workshop bench entry reviewed in cycle 9"
+  },
+  {
+    id: "wor-43-020",
+    label: "granite annex 20",
+    capacity: 76,
+    priority: 3,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 9"
+  },
+  {
+    id: "wor-43-021",
+    label: "central row 21",
+    capacity: 40,
+    priority: 5,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 5"
+  },
+  {
+    id: "wor-43-022",
+    label: "granite annex 22",
+    capacity: 68,
+    priority: 3,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-023",
+    label: "maple row 23",
+    capacity: 78,
+    priority: 5,
+    active: false,
+    notes: "central workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-024",
+    label: "granite hall 24",
+    capacity: 18,
+    priority: 4,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 9"
+  },
+  {
+    id: "wor-43-025",
+    label: "central hall 25",
+    capacity: 78,
+    priority: 1,
+    active: false,
+    notes: "central workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-026",
+    label: "northern row 26",
+    capacity: 56,
+    priority: 1,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 5"
+  },
+  {
+    id: "wor-43-027",
+    label: "northern annex 27",
+    capacity: 96,
+    priority: 2,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-028",
+    label: "central hall 28",
+    capacity: 36,
+    priority: 5,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 5"
+  },
+  {
+    id: "wor-43-029",
+    label: "granite annex 29",
+    capacity: 34,
+    priority: 2,
+    active: false,
+    notes: "central workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-030",
+    label: "northern row 30",
+    capacity: 12,
+    priority: 4,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 9"
+  },
+  {
+    id: "wor-43-031",
+    label: "central row 31",
+    capacity: 42,
+    priority: 3,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 5"
+  },
+  {
+    id: "wor-43-032",
+    label: "central hall 32",
+    capacity: 90,
+    priority: 5,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-033",
+    label: "central annex 33",
+    capacity: 70,
+    priority: 1,
+    active: false,
+    notes: "northern workshop bench entry reviewed in cycle 1"
+  },
+  {
+    id: "wor-43-034",
+    label: "central hall 34",
+    capacity: 50,
+    priority: 1,
+    active: false,
+    notes: "granite workshop bench entry reviewed in cycle 5"
+  }
+];
+
+export function findWorkshopBenchRecord43(id: string): WorkshopBenchRecord43 | undefined {
+  return workshopBenchRecords43.find((record) => record.id === id);
+}
+
+export function listWorkshopBenchActiveRecords43(): WorkshopBenchRecord43[] {
+  return workshopBenchRecords43.filter((record) => record.active);
+}
+
+export function totalWorkshopBenchCapacity43(): number {
+  return workshopBenchRecords43.reduce((sum, record) => sum + record.capacity, 0);
+}
+
+export function countWorkshopBenchByPriority43(priority: number): number {
+  return workshopBenchRecords43.filter((record) => record.priority === priority).length;
+}

package/benchmarks/projects/context-window-scaling-fixed-ts/src/tasks/auditRedaction.ts

@@ -0,0 +1,25 @@
+export interface AuditEntry {
+  actor: string;
+  action: string;
+  email?: string;
+  token?: string;
+}
+
+export function maskEmail(email: string): string {
+  const [local, domain] = email.split("@");
+  if (!local || !domain) {
+    return "[invalid-email]";
+  }
+  return `${local[0]}***@${domain}`;
+}
+
+export function redactAuditEntry(entry: AuditEntry): AuditEntry {
+  const redacted: AuditEntry = { actor: entry.actor, action: entry.action };
+  if (entry.email !== undefined) {
+    redacted.email = maskEmail(entry.email);
+  }
+  if (entry.token !== undefined) {
+    redacted.token = "[redacted]";
+  }
+  return redacted;
+}

package/benchmarks/projects/context-window-scaling-fixed-ts/src/tasks/inventoryReservation.ts

@@ -0,0 +1,26 @@
+export interface StockLevel {
+  sku: string;
+  onHand: number;
+  reserved: number;
+}
+
+export function availableStock(level: StockLevel): number {
+  return level.onHand - level.reserved;
+}
+
+export function reserveStock(level: StockLevel, quantity: number): StockLevel {
+  if (!Number.isInteger(quantity) || quantity <= 0) {
+    throw new Error("Reservation quantity must be a positive integer.");
+  }
+  if (quantity > availableStock(level)) {
+    throw new Error(`Insufficient stock for ${level.sku}.`);
+  }
+  return { ...level, reserved: level.reserved + quantity };
+}
+
+export function releaseReservation(level: StockLevel, quantity: number): StockLevel {
+  if (!Number.isInteger(quantity) || quantity <= 0 || quantity > level.reserved) {
+    throw new Error("Release quantity must be a positive integer no larger than the reserved amount.");
+  }
+  return { ...level, reserved: level.reserved - quantity };
+}

package/benchmarks/projects/context-window-scaling-fixed-ts/src/tasks/invoiceTotals.ts

@@ -0,0 +1,25 @@
+export interface InvoiceLine {
+  unitCents: number;
+  quantity: number;
+}
+
+export function roundToCents(value: number): number {
+  return Math.round(value + Number.EPSILON);
+}
+
+export function applyBulkDiscount(subtotalCents: number, totalQuantity: number): number {
+  if (totalQuantity >= 100) {
+    return roundToCents(subtotalCents * 0.9);
+  }
+  if (totalQuantity >= 20) {
+    return roundToCents(subtotalCents * 0.95);
+  }
+  return subtotalCents;
+}
+
+export function calculateInvoiceTotal(lines: readonly InvoiceLine[], taxRate: number): number {
+  const subtotal = lines.reduce((sum, line) => sum + line.unitCents * line.quantity, 0);
+  const quantity = lines.reduce((sum, line) => sum + line.quantity, 0);
+  const discounted = applyBulkDiscount(subtotal, quantity);
+  return discounted + roundToCents(discounted * taxRate);
+}

package/benchmarks/projects/context-window-scaling-fixed-ts/src/tasks/shippingQuote.ts

@@ -0,0 +1,32 @@
+export interface Parcel {
+  weightGrams: number;
+  destinationZone: "domestic" | "regional" | "international";
+}
+
+export interface WeightBand {
+  maxGrams: number;
+  baseCents: number;
+}
+
+export const weightBands: readonly WeightBand[] = [
+  { maxGrams: 500, baseCents: 450 },
+  { maxGrams: 2000, baseCents: 900 },
+  { maxGrams: 10000, baseCents: 1800 }
+];
+
+const zoneSurchargeCents = { domestic: 0, regional: 250, international: 1200 } as const;
+
+export function selectWeightBand(weightGrams: number): WeightBand {
+  if (!Number.isFinite(weightGrams) || weightGrams <= 0) {
+    throw new Error("Parcel weight must be a positive number of grams.");
+  }
+  const band = weightBands.find((candidate) => weightGrams <= candidate.maxGrams);
+  if (!band) {
+    throw new Error("Parcel exceeds the maximum supported weight.");
+  }
+  return band;
+}
+
+export function calculateShippingQuote(parcel: Parcel): number {
+  return selectWeightBand(parcel.weightGrams).baseCents + zoneSurchargeCents[parcel.destinationZone];
+}

package/benchmarks/projects/context-window-scaling-fixed-ts/tests/tasks/auditRedaction.test.ts

@@ -0,0 +1,17 @@
+import { describe, expect, it } from "vitest";
+import { maskEmail, redactAuditEntry } from "../../src/tasks/auditRedaction.js";
+
+describe("auditRedaction", () => {
+  it("masks the email local part and redacts tokens", () => {
+    expect(redactAuditEntry({ actor: "ana", action: "login", email: "ana@example.com", token: "abc" })).toEqual({
+      actor: "ana",
+      action: "login",
+      email: "a***@example.com",
+      token: "[redacted]"
+    });
+  });
+
+  it("flags malformed emails", () => {
+    expect(maskEmail("not-an-email")).toBe("[invalid-email]");
+  });
+});

package/benchmarks/projects/context-window-scaling-fixed-ts/tests/tasks/inventoryReservation.test.ts

@@ -0,0 +1,19 @@
+import { describe, expect, it } from "vitest";
+import { availableStock, releaseReservation, reserveStock } from "../../src/tasks/inventoryReservation.js";
+
+describe("inventoryReservation", () => {
+  it("reserves and releases stock without mutating the input", () => {
+    const level = { sku: "sku-1", onHand: 10, reserved: 2 };
+    const reserved = reserveStock(level, 3);
+    expect(reserved.reserved).toBe(5);
+    expect(level.reserved).toBe(2);
+    expect(availableStock(releaseReservation(reserved, 5))).toBe(10);
+  });
+
+  it("rejects over-reservation and invalid quantities", () => {
+    const level = { sku: "sku-1", onHand: 4, reserved: 1 };
+    expect(() => reserveStock(level, 4)).toThrow("Insufficient stock");
+    expect(() => reserveStock(level, 0)).toThrow("positive integer");
+    expect(() => releaseReservation(level, 2)).toThrow("no larger");
+  });
+});

package/benchmarks/projects/context-window-scaling-fixed-ts/tests/tasks/invoiceTotals.test.ts

@@ -0,0 +1,14 @@
+import { describe, expect, it } from "vitest";
+import { applyBulkDiscount, calculateInvoiceTotal } from "../../src/tasks/invoiceTotals.js";
+
+describe("invoiceTotals", () => {
+  it("applies tiered bulk discounts by quantity", () => {
+    expect(applyBulkDiscount(1000, 19)).toBe(1000);
+    expect(applyBulkDiscount(1000, 20)).toBe(950);
+    expect(applyBulkDiscount(1000, 100)).toBe(900);
+  });
+
+  it("adds rounded tax after the discount", () => {
+    expect(calculateInvoiceTotal([{ unitCents: 500, quantity: 2 }], 0.1)).toBe(1100);
+  });
+});

package/benchmarks/projects/context-window-scaling-fixed-ts/tests/tasks/shippingQuote.test.ts

@@ -0,0 +1,19 @@
+import { describe, expect, it } from "vitest";
+import { calculateShippingQuote, selectWeightBand } from "../../src/tasks/shippingQuote.js";
+
+describe("shippingQuote", () => {
+  it("selects the smallest band that holds the parcel", () => {
+    expect(selectWeightBand(500).baseCents).toBe(450);
+    expect(selectWeightBand(501).baseCents).toBe(900);
+  });
+
+  it("adds the zone surcharge to the band price", () => {
+    expect(calculateShippingQuote({ weightGrams: 300, destinationZone: "domestic" })).toBe(450);
+    expect(calculateShippingQuote({ weightGrams: 300, destinationZone: "international" })).toBe(1650);
+  });
+
+  it("rejects non-positive and oversize weights", () => {
+    expect(() => selectWeightBand(0)).toThrow("positive");
+    expect(() => selectWeightBand(10001)).toThrow("maximum");
+  });
+});

package/dist/src/cli/help.js

@@ -143,7 +143,10 @@ export function renderExperimentRunHelp() {
         "  --experiment <id>   Registered experiment plugin id (see \"my-dev-kit-lab experiment list\")",
         "",
         "Common options (all plugins):",
-        "  --target <path>                                    Local target project path (default: self)",
+        "  --target <path>                                    Local target project path (default: self); only for",
+        "                                                      plugins that support external targets. Self-only",
+        "                                                      plugins (for example incremental-change-staleness)",
+        "                                                      reject an explicit external --target.",
         "  --out <dir>                                         Output directory for experiment artifacts",
         "                                                      (default: beneath the workspace when installed)",
         "  --cases <path>                                      Path to the evaluation cases file",
@@ -153,10 +156,22 @@ export function renderExperimentRunHelp() {
         "  --case <ids>                                       Comma-separated case ids to run",
         "  --benchmark-project <ids>                          Comma-separated benchmark project ids to run",
         "",
+        "my-dev-kit command override (warm-index-reuse, incremental-change-staleness, and context-window-scaling):",
+        "  --kit-command <command>                            my-dev-kit command used by the selected experiment plugin",
+        "                                                      (defaults: warm-index-reuse and context-window-scaling use",
+        "                                                      npx @dailephd/my-dev-kit@latest;",
+        "                                                      incremental-change-staleness uses npx @dailephd/my-dev-kit@1.12.5)",
+        "",
+        "context-window-scaling only:",
+        "  --context-budgets <values>                          Comma-separated estimated-context-token budgets: 8k, 16k, 32k,",
+        "                                                      64k (8192/16384/32768/65536), or positive integers such as",
+        "                                                      12000 (default: 8k,16k,32k,64k). Duplicates after",
+        "                                                      normalization are rejected. Only --out, --case,",
+        "                                                      --context-budgets, and --kit-command are accepted; --case",
+        "                                                      selects from the bundled four-case catalog. Self target only.",
+        "                                                      Generate plots with \"plots generate --experiment <run dir>\".",
+        "",
         "warm-index-reuse only:",
-        "  --kit-command <command>                            my-dev-kit command used to build one index per benchmark",
-        "                                                      project and retrieve per task",
-        "                                                      (default: npx @dailephd/my-dev-kit@latest)",
         "                                                      Legacy/default mode uses the deterministic fake agent only;",
         "                                                      agent options below are not accepted by this plugin.",
         "  --campaign <preset>                                 Runs the selected real-agent campaign preset: codex-full,",
@@ -173,11 +188,13 @@ export function renderExperimentRunHelp() {
         "                                                      replaced by a context-size estimate. A campaign run",
         "                                                      automatically produces the report JSON/text/HTML, the",
         "                                                      four warm-index SVG plots, a best-effort report PNG",
-        "                                                      screenshot, and a gallery manifest/index. A missing",
-        "                                                      Playwright/browser runtime makes screenshot capture skip",
-        "                                                      without invalidating the campaign; an unexpected",
-        "                                                      screenshot failure makes the command unsuccessful after",
-        "                                                      preserving the report, plots, and gallery. Legacy",
+        "                                                      screenshot, and a gallery manifest/index. The automatic",
+        "                                                      campaign report screenshot is best-effort presentation",
+        "                                                      evidence: skipped or failed capture is reported explicitly",
+        "                                                      (failed remains failed), no PNG is fabricated, and",
+        "                                                      screenshot status alone does not fail an otherwise",
+        "                                                      successful campaign command. Core experiment, report,",
+        "                                                      plot, and gallery failures remain fatal. Legacy",
         "                                                      non-campaign warm-index runs retain their existing",
         "                                                      report-only behavior.",
         "",
@@ -283,7 +300,7 @@ export function renderPlotsGenerateHelp() {
         "Supported inputs:",
         "  - legacy controlled-experiment output directories",
         "  - warm-index-reuse plugin output directories (containing its report.json)",
-        "  Other plugin outputs are not plotted by this command.",
+        "  - context-window-scaling plugin output directories (containing context-window-scaling-execution.json)",
         "",
         "Required:",
         "  --experiment <dir>   Path to a supported experiment output directory",

package/dist/src/commands/runExperimentDescribeCommand.js

@@ -147,9 +147,12 @@ function describeExamples(plugin) {
     const screenshotOption = hasConfigField(plugin, "agents") ? " --no-screenshot" : "";
     const examples = [
         `my-dev-kit-lab experiment describe --experiment ${plugin.metadata.id}`,
-        `my-dev-kit-lab experiment run --experiment ${plugin.metadata.id}${agentOptions}`,
-        `my-dev-kit-lab experiment run --experiment ${plugin.metadata.id} --target "Z:\\Users\\newuser\\Projects\\my-dev-kit-v1"${agentOptions}${screenshotOption}`
+        `my-dev-kit-lab experiment run --experiment ${plugin.metadata.id}${agentOptions}`
     ];
+    // Only advertise an explicit --target example for plugins that support external targets.
+    if (plugin.metadata.supportedTargets.includes("external-local")) {
+        examples.push(`my-dev-kit-lab experiment run --experiment ${plugin.metadata.id} --target "Z:\\Users\\newuser\\Projects\\my-dev-kit-v1"${agentOptions}${screenshotOption}`);
+    }
     // Only advertise campaign examples for plugins that declare the campaignPreset config field.
     if (hasConfigField(plugin, "campaignPreset")) {
         examples.push(`my-dev-kit-lab experiment run --experiment ${plugin.metadata.id} --campaign codex-full --include-real-agents --out <dir>`, `my-dev-kit-lab experiment run --experiment ${plugin.metadata.id} --campaign claude-full --include-real-agents --case warm-medium-complete-idempotent --out <dir>`);