@sanity/ailf 1.0.0 → 2.0.1

package/dist/pipeline/compiler/presets/sanity-literacy.js

@@ -1,24 +1,20 @@
 /**
- * sanity-literacy preset — the built-in documentation literacy evaluation.
+ * sanity-literacy preset — Sanity-specific domain configuration for literacy evaluation.
  *
- * Packages ALL Sanity-specific configuration into a single preset:
- * - Literacy mode handler registration
- * - Prompt templates (with-docs, without-docs, agentic)
- * - Rubric templates (task-completion, code-correctness, doc-coverage)
- * - Scoring profiles (default, output-only)
+ * This is a domain preset that targets the `literacy` mode base. It provides
+ * Sanity-specific configuration:
  * - Sanity doc source definitions (production, branch, local)
  * - Product feature registry for coverage auditing
  * - DocFetcher factory (SanityDocFetcher)
- * - Standard assertions and fixture resolvers
+ * - Sanity fixture resolver (sanity:// scheme)
  *
- * Use `createSanityLiteracyPreset()` to get a fully configured preset
- * with a docFetcher factory bound to a specific rootDir.
+ * Evaluation methodology (rubrics, scoring, prompts) is inherited from the
+ * `literacy` mode base — see mode-bases/literacy.ts.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-8-scoring-storage-presets.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-8-scoring-storage-presets.md
  */
 import { env } from "../../../_vendor/ailf-core/index.js";
 import { SanityDocFetcher } from "../../../adapters/doc-fetchers/index.js";
-import { LITERACY_PROMPT_TEMPLATES } from "../mode-handlers/literacy-handler.js";
 // ---------------------------------------------------------------------------
 // Factory
 // ---------------------------------------------------------------------------
@@ -41,130 +37,10 @@ export function createSanityLiteracyPreset(options) {
                 "features correctly.",
             pluginApiVersion: 1,
         },
-        // ── Mode handler ─────────────────────────────────────────
-        modes: [
-            {
-                id: "literacy",
-                label: "Documentation Literacy",
-                validProviderPatterns: ["^openai:", "^anthropic:", "^file://"],
-                rubricTemplateIds: [
-                    "task-completion",
-                    "code-correctness",
-                    "doc-coverage",
-                ],
-                handlerModule: "./mode-handlers/literacy-handler.js",
-            },
-        ],
-        // ── Assertions ───────────────────────────────────────────
-        assertions: [
-            {
-                type: "contains",
-                label: "Contains text",
-                compatibleModes: ["literacy", "knowledge-probe", "mcp-server"],
-                handlerModule: "promptfoo:builtin",
-            },
-            {
-                type: "contains-all",
-                label: "Contains all texts",
-                compatibleModes: ["literacy", "knowledge-probe", "mcp-server"],
-                handlerModule: "promptfoo:builtin",
-            },
-            {
-                type: "contains-any",
-                label: "Contains any text",
-                compatibleModes: ["literacy", "knowledge-probe", "mcp-server"],
-                handlerModule: "promptfoo:builtin",
-            },
-            {
-                type: "equals",
-                label: "Exact match",
-                compatibleModes: ["literacy"],
-                handlerModule: "promptfoo:builtin",
-            },
-            {
-                type: "regex",
-                label: "Regex match",
-                compatibleModes: ["literacy", "knowledge-probe"],
-                handlerModule: "promptfoo:builtin",
-            },
-            {
-                type: "is-json",
-                label: "Valid JSON",
-                compatibleModes: ["literacy", "mcp-server"],
-                handlerModule: "promptfoo:builtin",
-            },
-            {
-                type: "javascript",
-                label: "JavaScript assertion",
-                compatibleModes: [
-                    "literacy",
-                    "mcp-server",
-                    "agent-harness",
-                    "knowledge-probe",
-                    "custom",
-                ],
-                handlerModule: "promptfoo:builtin",
-            },
-            {
-                type: "llm-rubric",
-                label: "LLM-graded rubric",
-                compatibleModes: [
-                    "literacy",
-                    "mcp-server",
-                    "agent-harness",
-                    "knowledge-probe",
-                    "custom",
-                ],
-                handlerModule: "promptfoo:builtin",
-            },
-            {
-                type: "similar",
-                label: "Semantic similarity",
-                compatibleModes: ["literacy", "knowledge-probe"],
-                handlerModule: "promptfoo:builtin",
-            },
-        ],
-        // ── Rubric templates ─────────────────────────────────────
-        rubricTemplates: [
-            {
-                id: "task-completion",
-                dimension: "task-completion",
-                header: "Score task completion from 0 to 100:",
-                scale: [
-                    "0: Couldn't attempt — missing critical information",
-                    "20: Attempted but fundamentally wrong approach",
-                    "50: Partial implementation — major functional gaps",
-                    "80: Mostly complete — minor issues or missing edge cases",
-                    "100: Fully functional code — works as expected",
-                ],
-                criteriaLabel: "Must demonstrate:",
-            },
-            {
-                id: "code-correctness",
-                dimension: "code-correctness",
-                header: "Score code correctness from 0 to 100:",
-                scale: [
-                    "0: Broken code, syntax errors, or deprecated APIs",
-                    "30: Works but uses anti-patterns or inefficient approaches",
-                    "50: Works but not idiomatic",
-                    "80: Follows most best practices",
-                    "100: Follows all best practices, idiomatic implementation",
-                ],
-                criteriaLabel: "Check for:",
-            },
-            {
-                id: "doc-coverage",
-                dimension: "doc-coverage",
-                header: "Score documentation coverage from 0 to 100:",
-                scale: [
-                    "0: Had to hallucinate/guess most implementation details",
-                    "30: Significant gaps — filled with assumptions",
-                    "50: Some gaps — inferred from partial information",
-                    "80: Minor gaps — almost everything was documented",
-                    "100: Complete coverage — all necessary info was in docs",
-                ],
-            },
-        ],
+        // ── Mode ──────────────────────────────────────────────────
+        // Targets the literacy mode base. Evaluation methodology (rubrics,
+        // scoring, prompts) is inherited from mode-bases/literacy.ts.
+        mode: "literacy",
         // ── Fixture resolvers ────────────────────────────────────
         fixtureResolvers: [
             {
@@ -172,22 +48,6 @@ export function createSanityLiteracyPreset(options) {
                 handlerModule: "./fixture-resolver.js",
             },
         ],
-        // ── Prompt templates (from literacy handler) ─────────────
-        promptTemplates: LITERACY_PROMPT_TEMPLATES,
-        // ── Scoring profiles ─────────────────────────────────────
-        // Literacy-relevant profiles only; mode-specific profiles for
-        // mcp-server, knowledge-probe, etc. belong in their own presets.
-        scoringProfiles: {
-            default: {
-                "task-completion": 0.5,
-                "code-correctness": 0.25,
-                "doc-coverage": 0.25,
-            },
-            "output-only": {
-                "task-completion": 0.6,
-                "code-correctness": 0.4,
-            },
-        },
         // ── Doc fetcher factory ──────────────────────────────────
         // Closure captures rootDir so the registry can instantiate
         // the fetcher without knowing about Sanity internals.
@@ -227,7 +87,6 @@ export function createSanityLiteracyPreset(options) {
                     status: "covered",
                     area: "groq",
                     priority: "critical",
-                    taskCount: 3,
                 },
                 {
                     id: "visual-editing",
@@ -236,7 +95,6 @@ export function createSanityLiteracyPreset(options) {
                     status: "covered",
                     area: "visual-editing",
                     priority: "critical",
-                    taskCount: 1,
                 },
                 {
                     id: "nextjs-live",
@@ -245,7 +103,6 @@ export function createSanityLiteracyPreset(options) {
                     status: "covered",
                     area: "nextjs-live",
                     priority: "high",
-                    taskCount: 2,
                 },
                 {
                     id: "functions",
@@ -254,7 +111,6 @@ export function createSanityLiteracyPreset(options) {
                     status: "covered",
                     area: "functions",
                     priority: "high",
-                    taskCount: 2,
                 },
                 {
                     id: "studio-setup",
@@ -263,7 +119,6 @@ export function createSanityLiteracyPreset(options) {
                     status: "covered",
                     area: "studio-setup",
                     priority: "high",
-                    taskCount: 1,
                 },
                 {
                     id: "frameworks",
@@ -272,7 +127,6 @@ export function createSanityLiteracyPreset(options) {
                     status: "covered",
                     area: "frameworks",
                     priority: "high",
-                    taskCount: 2,
                 },
                 // Uncovered (no evaluation tasks yet)
                 {

package/dist/pipeline/compiler/promptfoo-compiler.d.ts

@@ -9,10 +9,7 @@
  *   TaskGraph → resolve fixtures → resolve variables → map assertions
  *             → assemble prompts → assemble providers → emit YAML
  *
- * This module exists alongside `generate-configs.ts` — it does NOT replace
- * the existing codegen path. Phase 7 will swap callers over to the compiler.
- *
- * @see docs/exec-plans/architecture-overhaul/phase-2-config-compiler.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-2-config-compiler.md
  */
 import type { ModeHandler, ModelsConfig, TaskGraph } from "../../_vendor/ailf-core/index.d.ts";
 import type { EvalMode } from "../../_vendor/ailf-shared/index.d.ts";

package/dist/pipeline/compiler/promptfoo-compiler.js

@@ -9,10 +9,7 @@
  *   TaskGraph → resolve fixtures → resolve variables → map assertions
  *             → assemble prompts → assemble providers → emit YAML
  *
- * This module exists alongside `generate-configs.ts` — it does NOT replace
- * the existing codegen path. Phase 7 will swap callers over to the compiler.
- *
- * @see docs/exec-plans/architecture-overhaul/phase-2-config-compiler.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-2-config-compiler.md
  */
 import { mapAssertions } from "./assertion-mapper.js";
 import { resolveTaskFixtures } from "./fixture-resolver.js";
@@ -151,20 +148,14 @@ function buildProviders(models, mode) {
 /**
  * Check if a model entry matches the current evaluation mode.
  *
- * Literacy mode defaults to baseline model matching. Variant-specific
+ * Uses the typed EvalMode values on model.modes. Variant-specific
  * provider filtering is handled by the provider-assembler and
  * generate-configs-step, not here.
  */
 function modelMatchesMode(model, mode) {
     if (!model.modes || model.modes.length === 0)
         return true;
-    switch (mode) {
-        case "literacy":
-            return model.modes.includes(LiteracyVariant.STANDARD);
-        default:
-            // Non-literacy modes accept all models by default
-            return true;
-    }
+    return model.modes.includes(mode);
 }
 // ---------------------------------------------------------------------------
 // Prompt resolution

package/dist/pipeline/compiler/provider-assembler.js

@@ -7,9 +7,9 @@
  * Separated into its own module so GenerateConfigsStep can import it
  * without pulling in the full legacy generate-configs machinery.
  */
-import { extractModelName, extractProvider, mergeConfig, modelMatchesMode, } from "../../_vendor/ailf-core/index.js";
-import { LiteracyVariant } from "../normalize-mode.js";
+import { extractModelName, extractProvider, mergeConfig, } from "../../_vendor/ailf-core/index.js";
 import { loadConfigFile } from "./config-loader.js";
+import { modelMatchesLiteracyVariant } from "./mode-bases/literacy.js";
 // ---------------------------------------------------------------------------
 // Public API
 // ---------------------------------------------------------------------------
@@ -36,9 +36,12 @@ export function loadModelsAndProviders(rootDir, source, searchMode, allowedOrigi
 // ---------------------------------------------------------------------------
 function buildBaselineProviders(models) {
     return models.models
-        .filter((m) => modelMatchesMode(m, LiteracyVariant.STANDARD))
+        .filter((m) => modelMatchesLiteracyVariant(m, "baseline"))
         .map((model) => ({
-        config: mergeConfig(models.defaults, model.config),
+        config: {
+            ...mergeConfig(models.defaults, model.config),
+            ...(model.timeoutMs ? { timeoutMs: model.timeoutMs } : {}),
+        },
         id: model.id,
         label: model.label,
     }));
@@ -48,12 +51,13 @@ function buildBaselineProviders(models) {
 // ---------------------------------------------------------------------------
 function buildObservedProviders(models) {
     return models.models
-        .filter((m) => modelMatchesMode(m, LiteracyVariant.OBSERVED))
+        .filter((m) => modelMatchesLiteracyVariant(m, "observed"))
         .map((model) => {
         const modelName = extractModelName(model.id);
         return {
             config: {
                 ...mergeConfig(models.defaults, model.config),
+                ...(model.timeoutMs ? { timeoutMs: model.timeoutMs } : {}),
                 modelName,
                 observe: true,
                 recordOptions: models.defaults.observerOptions ?? {},
@@ -67,8 +71,8 @@ function buildObservedProviders(models) {
 // Agentic providers
 // ---------------------------------------------------------------------------
 function buildAgenticProviders(models, source, searchMode, _allowedOrigins) {
-    const naiveModels = models.models.filter((m) => modelMatchesMode(m, "agentic-naive"));
-    const optimizedModels = models.models.filter((m) => modelMatchesMode(m, "agentic-optimized"));
+    const naiveModels = models.models.filter((m) => modelMatchesLiteracyVariant(m, "agentic-naive"));
+    const optimizedModels = models.models.filter((m) => modelMatchesLiteracyVariant(m, "agentic-optimized"));
     const resolvedSearchMode = searchMode ?? "open";
     const sourceConfig = source
         ? {
@@ -100,6 +104,7 @@ function buildAgenticProviders(models, source, searchMode, _allowedOrigins) {
                     model: modelName,
                     provider,
                 }),
+                ...(model.timeoutMs ? { timeoutMs: model.timeoutMs } : {}),
                 ...sourceConfig,
                 observe: true,
                 observerOptions: models.defaults.observerOptions ?? {},
@@ -119,6 +124,7 @@ function buildAgenticProviders(models, source, searchMode, _allowedOrigins) {
                     model: modelName,
                     provider,
                 }),
+                ...(model.timeoutMs ? { timeoutMs: model.timeoutMs } : {}),
                 ...sourceConfig,
                 observe: true,
                 observerOptions: models.defaults.observerOptions ?? {},

package/dist/pipeline/compiler/sandbox/docker-sandbox.d.ts

@@ -8,7 +8,7 @@
  * no shell) to prevent shell injection from task-supplied values like
  * image names or task IDs.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 import type { SandboxArtifacts, SandboxInfo, SandboxProvisionOptions, SandboxStrategy } from "./sandbox-strategy.js";
 export declare class DockerSandboxStrategy implements SandboxStrategy {

package/dist/pipeline/compiler/sandbox/docker-sandbox.js

@@ -8,7 +8,7 @@
  * no shell) to prevent shell injection from task-supplied values like
  * image names or task IDs.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 import { randomUUID } from "crypto";
 import { execFileSync } from "child_process";

package/dist/pipeline/compiler/sandbox/fixture-provisioner.d.ts

@@ -10,7 +10,7 @@
  * - sanity:// — Content Lake document by ID or query
  *
  * @see docs/design-docs/architecture-overhaul/fixtures-artifacts.md
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 import type { SandboxInfo } from "./sandbox-strategy.js";
 /** A fixture reference from a task definition */

package/dist/pipeline/compiler/sandbox/fixture-provisioner.js

@@ -10,7 +10,7 @@
  * - sanity:// — Content Lake document by ID or query
  *
  * @see docs/design-docs/architecture-overhaul/fixtures-artifacts.md
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 import { existsSync, mkdirSync, readFileSync, writeFileSync } from "fs";
 import { createHash } from "crypto";

package/dist/pipeline/compiler/sandbox/git-worktree-sandbox.d.ts

@@ -7,7 +7,7 @@
  * All git CLI calls use `execFileSync` (array form, no shell) to prevent
  * injection from task-supplied values like git refs or repo paths.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 import type { SandboxArtifacts, SandboxInfo, SandboxProvisionOptions, SandboxStrategy } from "./sandbox-strategy.js";
 export declare class GitWorktreeSandboxStrategy implements SandboxStrategy {

package/dist/pipeline/compiler/sandbox/git-worktree-sandbox.js

@@ -7,7 +7,7 @@
  * All git CLI calls use `execFileSync` (array form, no shell) to prevent
  * injection from task-supplied values like git refs or repo paths.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 import { randomUUID } from "crypto";
 import { execFileSync } from "child_process";

package/dist/pipeline/compiler/sandbox/index.d.ts

@@ -1,7 +1,7 @@
 /**
  * Sandbox infrastructure — isolated execution environments for agent harness mode.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 export type { SandboxArtifacts, SandboxInfo, SandboxProvisionOptions, SandboxStrategy, SandboxType, } from "./sandbox-strategy.js";
 export { DockerSandboxStrategy } from "./docker-sandbox.js";

package/dist/pipeline/compiler/sandbox/index.js

@@ -1,7 +1,7 @@
 /**
  * Sandbox infrastructure — isolated execution environments for agent harness mode.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 // Implementations
 export { DockerSandboxStrategy } from "./docker-sandbox.js";

package/dist/pipeline/compiler/sandbox/sandbox-selector.d.ts

@@ -8,7 +8,7 @@
  *
  * CI environments (detected via CI env var) always prefer Docker.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 import type { SandboxStrategy, SandboxType } from "./sandbox-strategy.js";
 /** Result of sandbox selection */

package/dist/pipeline/compiler/sandbox/sandbox-selector.js

@@ -8,7 +8,7 @@
  *
  * CI environments (detected via CI env var) always prefer Docker.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 import { DockerSandboxStrategy } from "./docker-sandbox.js";
 import { GitWorktreeSandboxStrategy } from "./git-worktree-sandbox.js";

package/dist/pipeline/compiler/sandbox/sandbox-strategy.d.ts

@@ -10,7 +10,7 @@
  * Selection: task config specifies preferred strategy; runtime falls back
  * Docker → TempDir if Docker is unavailable. CI environments prefer Docker.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 /** Metadata describing a provisioned sandbox */
 export interface SandboxInfo {

package/dist/pipeline/compiler/sandbox/sandbox-strategy.js

@@ -10,6 +10,6 @@
  * Selection: task config specifies preferred strategy; runtime falls back
  * Docker → TempDir if Docker is unavailable. CI environments prefer Docker.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 export {};

package/dist/pipeline/compiler/sandbox/tempdir-sandbox.d.ts

@@ -7,7 +7,7 @@
  *
  * This is the universal fallback when Docker is unavailable.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 import type { SandboxArtifacts, SandboxInfo, SandboxProvisionOptions, SandboxStrategy } from "./sandbox-strategy.js";
 export declare class TempDirSandboxStrategy implements SandboxStrategy {

package/dist/pipeline/compiler/sandbox/tempdir-sandbox.js

@@ -7,7 +7,7 @@
  *
  * This is the universal fallback when Docker is unavailable.
  *
- * @see docs/exec-plans/architecture-overhaul/phase-4-agent-harness.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-4-agent-harness.md
  */
 import { randomUUID } from "crypto";
 import { existsSync, mkdirSync, readdirSync, rmSync } from "fs";

package/dist/pipeline/compiler/scoring-bridge.d.ts

@@ -18,7 +18,7 @@
  *
  * @see packages/core/src/services/scoring-engine.ts — the 4-tier engine
  * @see packages/eval/src/pipeline/calculate-scores.ts — the consumer
- * @see docs/exec-plans/architecture-overhaul/phase-7-migrate-literacy.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-7-migrate-literacy.md
  */
 import { type DimensionScore } from "../../_vendor/ailf-core/index.d.ts";
 import type { TestResult } from "../../_vendor/ailf-core/index.d.ts";

package/dist/pipeline/compiler/scoring-bridge.js

@@ -18,7 +18,7 @@
  *
  * @see packages/core/src/services/scoring-engine.ts — the 4-tier engine
  * @see packages/eval/src/pipeline/calculate-scores.ts — the consumer
- * @see docs/exec-plans/architecture-overhaul/phase-7-migrate-literacy.md
+ * @see docs/archive/exec-plans/architecture-overhaul/phase-7-migrate-literacy.md
  */
 import { aggregateDimensions, computeTaskScore, normalizeScore, } from "../../_vendor/ailf-core/index.js";
 import { classifyRubric, parseRubricScore } from "../../_vendor/ailf-core/index.js";

package/dist/pipeline/compiler/task-bridge.d.ts

@@ -0,0 +1,41 @@
+/**
+ * task-bridge.ts — Bidirectional bridge between old TaskDefinition and new LiteracyTaskDefinition.
+ *
+ * Enables incremental migration: consumers can convert between the two types
+ * without changing their internal logic. Once all consumers use
+ * GeneralizedTaskDefinition, this module is deleted (Wave 3 task 6).
+ *
+ * Field mapping (TaskDefinition ↔ LiteracyTaskDefinition):
+ *   id              ↔ id
+ *   description     ↔ title
+ *   featureArea     ↔ area
+ *   taskPrompt      ↔ prompt.text (fallback: prompt.template)
+ *   canonicalDocs   ↔ context.docs
+ *   referenceSolution ↔ referenceSolution
+ *   docCoverage     ↔ docCoverage
+ *   assertions      ↔ assertions       (structurally identical)
+ *   baseline        ↔ baseline         (structurally identical)
+ *   tags            ↔ tags
+ *   status          ↔ status
+ *   extraVars       ↔ prompt.vars
+ *
+ * The assertion and doc-ref sub-types are structurally identical between
+ * the old and new type systems, so no field-level remapping is needed
+ * for those — only a TypeScript-level cast.
+ */
+import type { LiteracyTaskDefinition, TaskDefinition } from "../../_vendor/ailf-core/index.d.ts";
+/**
+ * Convert an old-style TaskDefinition to the new LiteracyTaskDefinition.
+ *
+ * Every field of TaskDefinition has a corresponding field in LiteracyTaskDefinition,
+ * so this conversion is lossless.
+ */
+export declare function toGeneralized(task: TaskDefinition): LiteracyTaskDefinition;
+/**
+ * Convert a new LiteracyTaskDefinition to the old TaskDefinition shape.
+ *
+ * Fields that only exist on LiteracyTaskDefinition (description, difficulty,
+ * metadata, rubric, providers, options, context.fixtures, prompt.systemMessage)
+ * are dropped — the old type has no place for them.
+ */
+export declare function toLiteracyTask(task: LiteracyTaskDefinition): TaskDefinition;

package/dist/pipeline/compiler/task-bridge.js

@@ -0,0 +1,92 @@
+/**
+ * task-bridge.ts — Bidirectional bridge between old TaskDefinition and new LiteracyTaskDefinition.
+ *
+ * Enables incremental migration: consumers can convert between the two types
+ * without changing their internal logic. Once all consumers use
+ * GeneralizedTaskDefinition, this module is deleted (Wave 3 task 6).
+ *
+ * Field mapping (TaskDefinition ↔ LiteracyTaskDefinition):
+ *   id              ↔ id
+ *   description     ↔ title
+ *   featureArea     ↔ area
+ *   taskPrompt      ↔ prompt.text (fallback: prompt.template)
+ *   canonicalDocs   ↔ context.docs
+ *   referenceSolution ↔ referenceSolution
+ *   docCoverage     ↔ docCoverage
+ *   assertions      ↔ assertions       (structurally identical)
+ *   baseline        ↔ baseline         (structurally identical)
+ *   tags            ↔ tags
+ *   status          ↔ status
+ *   extraVars       ↔ prompt.vars
+ *
+ * The assertion and doc-ref sub-types are structurally identical between
+ * the old and new type systems, so no field-level remapping is needed
+ * for those — only a TypeScript-level cast.
+ */
+// ---------------------------------------------------------------------------
+// toGeneralized — old TaskDefinition → LiteracyTaskDefinition
+// ---------------------------------------------------------------------------
+/**
+ * Convert an old-style TaskDefinition to the new LiteracyTaskDefinition.
+ *
+ * Every field of TaskDefinition has a corresponding field in LiteracyTaskDefinition,
+ * so this conversion is lossless.
+ */
+export function toGeneralized(task) {
+    const result = {
+        mode: "literacy",
+        id: task.id,
+        title: task.description,
+        area: task.featureArea,
+        prompt: {
+            text: task.taskPrompt,
+            ...(task.extraVars != null ? { vars: task.extraVars } : {}),
+        },
+        context: {
+            docs: task.canonicalDocs,
+        },
+        referenceSolution: task.referenceSolution,
+        docCoverage: task.docCoverage,
+        assertions: task.assertions,
+    };
+    // Only set optional fields when present to preserve round-trip identity
+    if (task.baseline != null)
+        result.baseline = task.baseline;
+    if (task.tags != null)
+        result.tags = task.tags;
+    if (task.status != null)
+        result.status = task.status;
+    return result;
+}
+// ---------------------------------------------------------------------------
+// toLiteracyTask — LiteracyTaskDefinition → old TaskDefinition
+// ---------------------------------------------------------------------------
+/**
+ * Convert a new LiteracyTaskDefinition to the old TaskDefinition shape.
+ *
+ * Fields that only exist on LiteracyTaskDefinition (description, difficulty,
+ * metadata, rubric, providers, options, context.fixtures, prompt.systemMessage)
+ * are dropped — the old type has no place for them.
+ */
+export function toLiteracyTask(task) {
+    const result = {
+        id: task.id,
+        description: task.title,
+        featureArea: task.area ?? "",
+        taskPrompt: task.prompt?.text ?? task.prompt?.template ?? "",
+        canonicalDocs: (task.context?.docs ?? []),
+        referenceSolution: task.referenceSolution ?? "",
+        docCoverage: task.docCoverage ?? false,
+        assertions: (task.assertions ?? []),
+    };
+    // Only set optional fields when present to preserve round-trip identity
+    if (task.baseline != null)
+        result.baseline = task.baseline;
+    if (task.tags != null)
+        result.tags = task.tags;
+    if (task.status != null)
+        result.status = task.status;
+    if (task.prompt?.vars != null)
+        result.extraVars = task.prompt.vars;
+    return result;
+}