@sanity/ailf 2.0.0 → 2.0.1

package/dist/_vendor/ailf-core/examples/index.js

@@ -2,7 +2,8 @@
  * src/examples/index.ts — Generated example data.
  *
  * DO NOT EDIT — this file is generated by scripts/generate-examples.ts
- * from the YAML files in packages/core/examples/.
+ * from TypeScript task files in packages/core/examples/tasks/
+ * and YAML config files in packages/core/examples/.
  *
  * To regenerate: pnpm generate-examples
  */
@@ -145,26 +146,29 @@ export const ailfConfigYaml = "# ───────────────
 /** Parsed task data for example-groq-blog-listing (JSON-safe) */
 export const exampleGroqBlogListingData = [
     {
+        "mode": "literacy",
         "id": "example-groq-blog-listing",
-        "description": "Example — Blog listing with GROQ queries",
-        "featureArea": "groq",
-        "canonicalDocs": [
-            {
-                "slug": "groq-introduction",
-                "reason": "Core GROQ syntax and query language reference"
-            },
-            {
-                "slug": "how-queries-work",
-                "reason": "Query execution model and best practices"
-            }
-        ],
+        "title": "Blog listing with GROQ queries",
+        "description": "Example — tests GROQ blog listing implementation",
+        "area": "groq",
+        "context": {
+            "docs": [
+                {
+                    "slug": "groq-introduction",
+                    "reason": "Core GROQ syntax and query language reference"
+                },
+                {
+                    "slug": "how-queries-work",
+                    "reason": "Query execution model and best practices"
+                }
+            ]
+        },
         "docCoverage": true,
         "referenceSolution": "canonical/example-groq-blog-listing.ts",
-        "vars": {
-            "task": "Create a Next.js page component that lists blog posts from Sanity\nusing GROQ. The page should display the title, slug, and published\ndate for each post, sorted by most recent first. Use the Sanity\nclient to fetch data.\n",
-            "docs": ""
+        "prompt": {
+            "text": "Create a Next.js page component that lists blog posts from Sanity\nusing GROQ. The page should display the title, slug, and published\ndate for each post, sorted by most recent first. Use the Sanity\nclient to fetch data."
         },
-        "assert": [
+        "assertions": [
             {
                 "type": "llm-rubric",
                 "template": "task-completion",
@@ -190,32 +194,37 @@ export const exampleGroqBlogListingData = [
         "status": "draft"
     }
 ];
-/** Raw YAML string for example-groq-blog-listing (preserves comments) */
-export const exampleGroqBlogListingYaml = "# ──────────────────────────────────────────────────────────────────────\n# Example Task: Blog listing with GROQ queries\n# ──────────────────────────────────────────────────────────────────────\n#\n# This is a starter template — edit it for your own documentation.\n# Each task evaluates whether an AI coding agent can implement a feature\n# using your docs as context. Delete this file or replace it entirely.\n#\n# This example task ships as a DRAFT so it does not run in production\n#   evaluations automatically. To activate it, change status to \"active\"\n# or remove the status line entirely (defaults to active).\n#\n# Full field reference:\n#   https://github.com/sanity-labs/ai-literacy-framework/blob/main/docs/CONTRIBUTING_TASKS.md\n# ──────────────────────────────────────────────────────────────────────\n\n# Unique identifier — lowercase alphanumeric with hyphens.\n# Must be unique across all task files in .ailf/tasks/.\n- id: example-groq-blog-listing\n\n  # Short human-readable summary. Shown in score tables and reports.\n  description: \"Example — Blog listing with GROQ queries\"\n\n  # Feature area this task belongs to. Tasks with the same area are\n  # grouped together in score summaries. Use a short kebab-case name.\n  featureArea: groq\n\n  # Gold-standard documentation articles for this task. The pipeline\n  # fetches these from Sanity and injects them into the prompt for\n  # baseline evaluation. Each entry needs:\n  #   slug   — the article's URL slug in your docs site\n  #   reason — why this doc is relevant (helps with auditing)\n  #\n  # This example uses slug-based references — the simplest form.\n  # See the other example tasks for path, id, and perspective references.\n  canonicalDocs:\n    - slug: groq-introduction\n      reason: \"Core GROQ syntax and query language reference\"\n    - slug: how-queries-work\n      reason: \"Query execution model and best practices\"\n\n  # When true, the pipeline auto-generates an additional rubric that\n  # checks whether the LLM's response actually used the provided docs.\n  docCoverage: true\n\n  # Path to a gold-standard implementation, relative to canonical/.\n  # The grader uses this as a reference when scoring code correctness.\n  referenceSolution: canonical/example-groq-blog-listing.ts\n\n  # vars.task — the implementation prompt given to the LLM.\n  # Write this as if you're asking a developer to build the feature.\n  # Be specific about requirements so the grader can evaluate clearly.\n  #\n  # vars.docs — leave empty (\"\"). The pipeline fills this in:\n  #   • Gold variant: injected with canonical doc content\n  #   • Baseline variant: left empty (tests model knowledge alone)\n  vars:\n    task: |\n      Create a Next.js page component that lists blog posts from Sanity\n      using GROQ. The page should display the title, slug, and published\n      date for each post, sorted by most recent first. Use the Sanity\n      client to fetch data.\n    docs: \"\"\n\n  # Grading assertions — how the LLM's response is scored.\n  #\n  # \"llm-rubric\" assertions use a grader LLM to score against criteria.\n  # The \"template\" references a rubric from config/rubrics.yaml.\n  # The \"criteria\" are task-specific bullets injected into the template.\n  #\n  # Available templates:\n  #   task-completion   — did the LLM implement the feature? (weight: 0.50)\n  #   code-correctness  — is the code idiomatic and correct? (weight: 0.25)\n  #\n  # You can also use value-based assertions:\n  #   - type: contains\n  #     value: \"client.fetch\"\n  #   - type: contains-any\n  #     value: [\"createClient\", \"sanityClient\"]\n  assert:\n    - type: llm-rubric\n      template: task-completion\n      criteria:\n        - \"Uses the groq tagged template literal\"\n        - \"Fetches blog posts with title, slug, and publishedAt fields\"\n        - \"Orders results by publishedAt in descending order\"\n\n    - type: llm-rubric\n      template: code-correctness\n      criteria:\n        - \"Uses createClient from @sanity/client or next-sanity\"\n        - \"Exports a valid Next.js page component\"\n\n  # Baseline variant configuration.\n  #   enabled — set to false to skip this task entirely\n  #   rubric  — \"full\" (default), \"abbreviated\" (faster), or \"none\"\n  baseline:\n    enabled: true\n    rubric: full\n\n  # Example tasks ship as drafts so they don't run in production evals.\n  # Change to \"active\" (or remove this line) to activate.\n  status: draft\n";
+/** TypeScript task template for example-groq-blog-listing */
+export const exampleGroqBlogListingTs = "/**\n * Example Task: Blog listing with GROQ queries.\n *\n * This is a starter template — edit it for your own documentation.\n * Each task evaluates whether an AI coding agent can implement a feature\n * using your docs as context. Delete this file or replace it entirely.\n *\n * This example task ships as a DRAFT so it does not run in production\n * evaluations automatically. To activate it, change status to \"active\"\n * or remove the status field entirely (defaults to active).\n *\n * Full field reference:\n *   https://github.com/sanity-labs/ai-literacy-framework/blob/main/docs/contributing-tasks.md\n */\n\nimport { defineTask } from \"@sanity/ailf-core\"\n\nexport default defineTask({\n  // ── Mode ────────────────────────────────────────────────────────────\n  // \"literacy\" tests whether AI coding tools can implement features\n  // using your docs as context. Other modes: \"mcp-server\",\n  // \"knowledge-probe\", \"agent-harness\", \"custom\".\n  mode: \"literacy\",\n\n  // ── Identity ────────────────────────────────────────────────────────\n  // Unique identifier — lowercase alphanumeric with hyphens.\n  // Must be unique across all task files in .ailf/tasks/.\n  id: \"example-groq-blog-listing\",\n  title: \"Blog listing with GROQ queries\",\n  description: \"Example — tests GROQ blog listing implementation\",\n\n  // Feature area this task belongs to. Tasks with the same area are\n  // grouped together in score summaries.\n  area: \"groq\",\n\n  // ── Documentation context ───────────────────────────────────────────\n  // Canonical doc references for this task. The pipeline fetches these\n  // from Sanity and injects them into the prompt for baseline evaluation.\n  //\n  // This example uses slug-based references — the simplest form.\n  // See the other example tasks for path, id, and perspective references.\n  context: {\n    docs: [\n      {\n        slug: \"groq-introduction\",\n        reason: \"Core GROQ syntax and query language reference\",\n      },\n      {\n        slug: \"how-queries-work\",\n        reason: \"Query execution model and best practices\",\n      },\n    ],\n  },\n\n  // When true, the pipeline auto-generates an additional rubric that\n  // checks whether the LLM's response actually used the provided docs.\n  docCoverage: true,\n\n  // Path to a gold-standard implementation, relative to canonical/.\n  // The grader uses this as a reference when scoring code correctness.\n  referenceSolution: \"canonical/example-groq-blog-listing.ts\",\n\n  // ── Prompt ──────────────────────────────────────────────────────────\n  // prompt.text — the implementation prompt given to the LLM.\n  // Write this as if you're asking a developer to build the feature.\n  // Be specific about requirements so the grader can evaluate clearly.\n  prompt: {\n    text: `Create a Next.js page component that lists blog posts from Sanity\nusing GROQ. The page should display the title, slug, and published\ndate for each post, sorted by most recent first. Use the Sanity\nclient to fetch data.`,\n  },\n\n  // ── Assertions ──────────────────────────────────────────────────────\n  // Grading assertions — how the LLM's response is scored.\n  //\n  // \"llm-rubric\" assertions use a grader LLM to score against criteria.\n  // The \"template\" references a rubric template (e.g. task-completion).\n  //\n  // Available templates:\n  //   task-completion   — did the LLM implement the feature? (weight: 0.50)\n  //   code-correctness  — is the code idiomatic and correct? (weight: 0.25)\n  //\n  // You can also use value-based assertions:\n  //   { type: \"contains\", value: \"client.fetch\" }\n  //   { type: \"contains-any\", value: [\"createClient\", \"sanityClient\"] }\n  assertions: [\n    {\n      type: \"llm-rubric\",\n      template: \"task-completion\",\n      criteria: [\n        \"Uses the groq tagged template literal\",\n        \"Fetches blog posts with title, slug, and publishedAt fields\",\n        \"Orders results by publishedAt in descending order\",\n      ],\n    },\n    {\n      type: \"llm-rubric\",\n      template: \"code-correctness\",\n      criteria: [\n        \"Uses createClient from @sanity/client or next-sanity\",\n        \"Exports a valid Next.js page component\",\n      ],\n    },\n  ],\n\n  // ── Baseline variant ────────────────────────────────────────────────\n  //   enabled — set to false to skip this task entirely\n  //   rubric  — \"full\" (default), \"abbreviated\" (faster), or \"none\"\n  baseline: {\n    enabled: true,\n    rubric: \"full\",\n  },\n\n  // Example tasks ship as drafts so they don't run in production evals.\n  // Change to \"active\" (or remove this field) to activate.\n  status: \"draft\",\n})\n";
+/** Generated YAML for example-groq-blog-listing (from parsed TS data) */
+export const exampleGroqBlogListingYaml = "- mode: literacy\n  id: example-groq-blog-listing\n  title: Blog listing with GROQ queries\n  description: Example — tests GROQ blog listing implementation\n  area: groq\n  context:\n    docs:\n      - slug: groq-introduction\n        reason: Core GROQ syntax and query language reference\n      - slug: how-queries-work\n        reason: Query execution model and best practices\n  docCoverage: true\n  referenceSolution: canonical/example-groq-blog-listing.ts\n  prompt:\n    text: |-\n      Create a Next.js page component that lists blog posts from Sanity\n      using GROQ. The page should display the title, slug, and published\n      date for each post, sorted by most recent first. Use the Sanity\n      client to fetch data.\n  assertions:\n    - type: llm-rubric\n      template: task-completion\n      criteria:\n        - Uses the groq tagged template literal\n        - Fetches blog posts with title, slug, and publishedAt fields\n        - Orders results by publishedAt in descending order\n    - type: llm-rubric\n      template: code-correctness\n      criteria:\n        - Uses createClient from @sanity/client or next-sanity\n        - Exports a valid Next.js page component\n  baseline:\n    enabled: true\n    rubric: full\n  status: draft\n";
 /** Parsed task data for example-id-based-ref (JSON-safe) */
 export const exampleIdBasedRefData = [
     {
+        "mode": "literacy",
         "id": "example-id-based-ref",
-        "description": "Example — GROQ feature support (ID-based doc references)",
-        "featureArea": "groq",
-        "canonicalDocs": [
-            {
-                "id": "0ba88f1b-d1a7-418a-9267-2e343d01886a",
-                "slug": "groq-feature-support-by-context",
-                "reason": "GROQ feature support across different Sanity contexts"
-            },
-            {
-                "id": "5b9c2863-ef01-4565-af8e-ee54e081ee74",
-                "slug": "custom-groq-functions",
-                "reason": "Custom GROQ functions and pipelines"
-            }
-        ],
+        "title": "GROQ feature support (ID-based doc references)",
+        "description": "Example — demonstrates ID-based canonical doc references",
+        "area": "groq",
+        "context": {
+            "docs": [
+                {
+                    "id": "0ba88f1b-d1a7-418a-9267-2e343d01886a",
+                    "slug": "groq-feature-support-by-context",
+                    "reason": "GROQ feature support across different Sanity contexts"
+                },
+                {
+                    "id": "5b9c2863-ef01-4565-af8e-ee54e081ee74",
+                    "slug": "custom-groq-functions",
+                    "reason": "Custom GROQ functions and pipelines"
+                }
+            ]
+        },
         "docCoverage": true,
-        "vars": {
-            "task": "Explain how GROQ is used across different Sanity contexts.\nCover the following:\n1. Which GROQ features are available in each context (API queries,\n   webhooks, custom functions, access control)\n2. How to create and use custom GROQ functions\n3. Any differences in GROQ support between contexts\nProvide examples demonstrating context-specific GROQ patterns.\n",
-            "docs": ""
+        "prompt": {
+            "text": "Explain how GROQ is used across different Sanity contexts.\nCover the following:\n1. Which GROQ features are available in each context (API queries,\n   webhooks, custom functions, access control)\n2. How to create and use custom GROQ functions\n3. Any differences in GROQ support between contexts\nProvide examples demonstrating context-specific GROQ patterns."
         },
-        "assert": [
+        "assertions": [
             {
                 "type": "llm-rubric",
                 "template": "task-completion",
@@ -241,30 +250,100 @@ export const exampleIdBasedRefData = [
         "status": "draft"
     }
 ];
-/** Raw YAML string for example-id-based-ref (preserves comments) */
-export const exampleIdBasedRefYaml = "# ──────────────────────────────────────────────────────────────────────\n# Example Task: Document ID-based canonical doc references\n# ──────────────────────────────────────────────────────────────────────\n#\n# Demonstrates using `id` to reference canonical documentation by\n# Sanity document `_id`. This is useful for:\n#   - Draft documents that don't have a stable slug yet\n#   - Programmatic references from imports or migrations\n#   - Documents where you know the _id but not the slug\n#\n# The `id` ref type can also carry optional `slug` and `path` fields\n# as human-readable annotations — these are NOT used for resolution,\n# only for display in logs and reports.\n#\n# This example task ships as a DRAFT so it does not run in production\n#   evaluations automatically. To activate it, change status to \"active\"\n# or remove the status line entirely (defaults to active).\n#\n# @see docs/design-docs/canonical-doc-resolution.md\n# ──────────────────────────────────────────────────────────────────────\n\n- id: example-id-based-ref\n  description: \"Example — GROQ feature support (ID-based doc references)\"\n\n  featureArea: groq\n\n  # ID-based canonical doc references.\n  #\n  # Use the Sanity document _id to reference articles directly.\n  # Optional slug/path annotations help humans reading the YAML\n  # but are NOT used for resolution — only the `id` field matters.\n  #\n  # These IDs reference real articles in the Sanity docs (next dataset):\n  #   0ba88f1b... = \"GROQ feature support across Sanity\"\n  #   5b9c2863... = \"Custom GROQ functions\"\n  canonicalDocs:\n    - id: \"0ba88f1b-d1a7-418a-9267-2e343d01886a\"\n      slug: groq-feature-support-by-context # annotation only — not used for resolution\n      reason: \"GROQ feature support across different Sanity contexts\"\n    - id: \"5b9c2863-ef01-4565-af8e-ee54e081ee74\"\n      slug: custom-groq-functions # annotation only — not used for resolution\n      reason: \"Custom GROQ functions and pipelines\"\n\n  docCoverage: true\n\n  vars:\n    task: |\n      Explain how GROQ is used across different Sanity contexts.\n      Cover the following:\n      1. Which GROQ features are available in each context (API queries,\n         webhooks, custom functions, access control)\n      2. How to create and use custom GROQ functions\n      3. Any differences in GROQ support between contexts\n      Provide examples demonstrating context-specific GROQ patterns.\n    docs: \"\"\n\n  assert:\n    - type: llm-rubric\n      template: task-completion\n      criteria:\n        - \"Explains GROQ availability across different Sanity contexts\"\n        - \"Describes custom GROQ function creation and usage\"\n        - \"Notes differences in GROQ support between contexts\"\n\n    - type: llm-rubric\n      template: code-correctness\n      criteria:\n        - \"GROQ examples use valid syntax\"\n        - \"Custom function examples follow the correct API pattern\"\n\n  baseline:\n    enabled: true\n    rubric: full\n\n  # Example tasks ship as drafts so they don't run in production evals.\n  # Change to \"active\" (or remove this line) to activate.\n  status: draft\n";
-/** Parsed task data for example-path-based-ref (JSON-safe) */
-export const examplePathBasedRefData = [
+/** TypeScript task template for example-id-based-ref */
+export const exampleIdBasedRefTs = "/**\n * Example Task: Document ID-based canonical doc references.\n *\n * Demonstrates using `id` to reference canonical documentation by\n * Sanity document `_id`. This is useful for:\n *   - Draft documents that don't have a stable slug yet\n *   - Programmatic references from imports or migrations\n *   - Documents where you know the _id but not the slug\n *\n * The `id` ref type can also carry optional `slug` and `path` fields\n * as human-readable annotations — these are NOT used for resolution,\n * only for display in logs and reports.\n *\n * This example task ships as a DRAFT so it does not run in production\n * evaluations automatically. To activate it, change status to \"active\"\n * or remove the status field entirely (defaults to active).\n *\n * @see docs/design-docs/canonical-doc-resolution.md\n */\n\nimport { defineTask } from \"@sanity/ailf-core\"\n\nexport default defineTask({\n  mode: \"literacy\",\n  id: \"example-id-based-ref\",\n  title: \"GROQ feature support (ID-based doc references)\",\n  description: \"Example — demonstrates ID-based canonical doc references\",\n\n  area: \"groq\",\n\n  // ID-based canonical doc references.\n  //\n  // Use the Sanity document _id to reference articles directly.\n  // Optional slug/path annotations help humans reading the file\n  // but are NOT used for resolution — only the `id` field matters.\n  //\n  // These IDs reference real articles in the Sanity docs (next dataset):\n  //   0ba88f1b... = \"GROQ feature support across Sanity\"\n  //   5b9c2863... = \"Custom GROQ functions\"\n  context: {\n    docs: [\n      {\n        id: \"0ba88f1b-d1a7-418a-9267-2e343d01886a\",\n        slug: \"groq-feature-support-by-context\", // annotation only\n        reason: \"GROQ feature support across different Sanity contexts\",\n      },\n      {\n        id: \"5b9c2863-ef01-4565-af8e-ee54e081ee74\",\n        slug: \"custom-groq-functions\", // annotation only\n        reason: \"Custom GROQ functions and pipelines\",\n      },\n    ],\n  },\n\n  docCoverage: true,\n\n  prompt: {\n    text: `Explain how GROQ is used across different Sanity contexts.\nCover the following:\n1. Which GROQ features are available in each context (API queries,\n   webhooks, custom functions, access control)\n2. How to create and use custom GROQ functions\n3. Any differences in GROQ support between contexts\nProvide examples demonstrating context-specific GROQ patterns.`,\n  },\n\n  assertions: [\n    {\n      type: \"llm-rubric\",\n      template: \"task-completion\",\n      criteria: [\n        \"Explains GROQ availability across different Sanity contexts\",\n        \"Describes custom GROQ function creation and usage\",\n        \"Notes differences in GROQ support between contexts\",\n      ],\n    },\n    {\n      type: \"llm-rubric\",\n      template: \"code-correctness\",\n      criteria: [\n        \"GROQ examples use valid syntax\",\n        \"Custom function examples follow the correct API pattern\",\n      ],\n    },\n  ],\n\n  baseline: { enabled: true, rubric: \"full\" },\n  status: \"draft\",\n})\n";
+/** Generated YAML for example-id-based-ref (from parsed TS data) */
+export const exampleIdBasedRefYaml = "- mode: literacy\n  id: example-id-based-ref\n  title: GROQ feature support (ID-based doc references)\n  description: Example — demonstrates ID-based canonical doc references\n  area: groq\n  context:\n    docs:\n      - id: 0ba88f1b-d1a7-418a-9267-2e343d01886a\n        slug: groq-feature-support-by-context\n        reason: GROQ feature support across different Sanity contexts\n      - id: 5b9c2863-ef01-4565-af8e-ee54e081ee74\n        slug: custom-groq-functions\n        reason: Custom GROQ functions and pipelines\n  docCoverage: true\n  prompt:\n    text: |-\n      Explain how GROQ is used across different Sanity contexts.\n      Cover the following:\n      1. Which GROQ features are available in each context (API queries,\n         webhooks, custom functions, access control)\n      2. How to create and use custom GROQ functions\n      3. Any differences in GROQ support between contexts\n      Provide examples demonstrating context-specific GROQ patterns.\n  assertions:\n    - type: llm-rubric\n      template: task-completion\n      criteria:\n        - Explains GROQ availability across different Sanity contexts\n        - Describes custom GROQ function creation and usage\n        - Notes differences in GROQ support between contexts\n    - type: llm-rubric\n      template: code-correctness\n      criteria:\n        - GROQ examples use valid syntax\n        - Custom function examples follow the correct API pattern\n  baseline:\n    enabled: true\n    rubric: full\n  status: draft\n";
+/** Parsed task data for example-knowledge-probe (JSON-safe) */
+export const exampleKnowledgeProbeData = [
     {
-        "id": "example-path-based-ref",
-        "description": "Example — GROQ mutations (path-based doc references)",
-        "featureArea": "groq",
-        "canonicalDocs": [
+        "mode": "knowledge-probe",
+        "id": "example-knowledge-probe",
+        "title": "Model knowledge of GROQ syntax",
+        "description": "Example — probes baseline model knowledge (draft)",
+        "area": "groq",
+        "prompt": {
+            "text": "Explain the GROQ query language used by Sanity. Cover:\n1. Basic query syntax and projections\n2. How to filter and sort results\n3. Common patterns for fetching related documents\nProvide working code examples."
+        },
+        "assertions": [
             {
-                "path": "content-lake/mutations-introduction",
-                "reason": "Introduction to document mutations in the Content Lake"
-            },
+                "type": "llm-rubric",
+                "template": "task-completion",
+                "criteria": [
+                    "Demonstrates understanding of GROQ query syntax",
+                    "Shows filtering and projection patterns",
+                    "Code examples use valid GROQ syntax"
+                ]
+            }
+        ],
+        "status": "draft"
+    }
+];
+/** TypeScript task template for example-knowledge-probe */
+export const exampleKnowledgeProbeTs = "/**\n * Example Task: Knowledge probe baseline (DRAFT).\n *\n * Tests what the model knows about a topic without providing documentation.\n * Used to establish a baseline for comparison with literacy evaluations.\n * This task is a DRAFT — it won't run unless activated or explicitly targeted.\n *\n * To activate: change status to \"active\" or remove the status field.\n */\n\nimport { defineTask } from \"@sanity/ailf-core\"\n\nexport default defineTask({\n  mode: \"knowledge-probe\",\n  id: \"example-knowledge-probe\",\n  title: \"Model knowledge of GROQ syntax\",\n  description: \"Example — probes baseline model knowledge (draft)\",\n  area: \"groq\",\n\n  prompt: {\n    text: `Explain the GROQ query language used by Sanity. Cover:\n1. Basic query syntax and projections\n2. How to filter and sort results\n3. Common patterns for fetching related documents\nProvide working code examples.`,\n  },\n\n  assertions: [\n    {\n      type: \"llm-rubric\",\n      template: \"task-completion\",\n      criteria: [\n        \"Demonstrates understanding of GROQ query syntax\",\n        \"Shows filtering and projection patterns\",\n        \"Code examples use valid GROQ syntax\",\n      ],\n    },\n  ],\n\n  status: \"draft\",\n})\n";
+/** Generated YAML for example-knowledge-probe (from parsed TS data) */
+export const exampleKnowledgeProbeYaml = "- mode: knowledge-probe\n  id: example-knowledge-probe\n  title: Model knowledge of GROQ syntax\n  description: Example — probes baseline model knowledge (draft)\n  area: groq\n  prompt:\n    text: |-\n      Explain the GROQ query language used by Sanity. Cover:\n      1. Basic query syntax and projections\n      2. How to filter and sort results\n      3. Common patterns for fetching related documents\n      Provide working code examples.\n  assertions:\n    - type: llm-rubric\n      template: task-completion\n      criteria:\n        - Demonstrates understanding of GROQ query syntax\n        - Shows filtering and projection patterns\n        - Code examples use valid GROQ syntax\n  status: draft\n";
+/** Parsed task data for example-mcp-tool-usage (JSON-safe) */
+export const exampleMcpToolUsageData = [
+    {
+        "mode": "mcp-server",
+        "id": "example-mcp-tool-usage",
+        "title": "MCP tool discovery and invocation",
+        "description": "Example — tests MCP server tool-use (draft)",
+        "area": "mcp",
+        "serverConfig": {
+            "transport": "streamable-http",
+            "url": "https://your-mcp-server.example.com",
+            "headers": {
+                "Authorization": "Bearer {{env.MCP_AUTH_TOKEN}}"
+            }
+        },
+        "prompt": {
+            "text": "Use the available MCP tools to complete the task.\nReplace this prompt with instructions specific to your MCP server."
+        },
+        "assertions": [
             {
-                "path": "content-lake/documents",
-                "reason": "Document structure and types (Content Lake, not CLI reference)"
+                "type": "llm-rubric",
+                "template": "mcp-output-correctness",
+                "criteria": [
+                    "Correctly discovers and selects the appropriate tool",
+                    "Passes valid arguments to the tool",
+                    "Interprets the tool response coherently"
+                ]
             }
         ],
+        "status": "draft"
+    }
+];
+/** TypeScript task template for example-mcp-tool-usage */
+export const exampleMcpToolUsageTs = "/**\n * Example Task: MCP Server tool-use evaluation (DRAFT).\n *\n * Tests whether an LLM can correctly discover and invoke tools from\n * an MCP server. Replace the placeholder serverConfig with your own\n * MCP server's URL and authentication details.\n *\n * Transports:\n *   - \"streamable-http\" / \"sse\" — remote servers (set url + optional headers)\n *   - \"stdio\" — local process (set command instead of url)\n *\n * Authentication:\n *   - `headers` — send arbitrary HTTP headers (e.g., Authorization)\n *   - `auth`    — structured auth config (bearer, basic, api_key, oauth)\n *   Values support {{env.VAR}} syntax so secrets stay out of source control.\n *\n * This task is a DRAFT — it won't run unless activated or explicitly targeted.\n * To activate: change status to \"active\" or remove the status field.\n */\n\nimport { defineTask } from \"@sanity/ailf-core\"\n\nexport default defineTask({\n  mode: \"mcp-server\",\n  id: \"example-mcp-tool-usage\",\n  title: \"MCP tool discovery and invocation\",\n  description: \"Example — tests MCP server tool-use (draft)\",\n  area: \"mcp\",\n\n  // ── Server configuration ────────────────────────────────────\n  // Replace the URL and headers below with your MCP server's details.\n  //\n  // For a local stdio server, use:\n  //   transport: \"stdio\",\n  //   command: \"node dist/my-mcp-server.js\",\n  serverConfig: {\n    transport: \"streamable-http\",\n    url: \"https://your-mcp-server.example.com\",\n    headers: {\n      Authorization: \"Bearer {{env.MCP_AUTH_TOKEN}}\",\n    },\n  },\n\n  // ── Capabilities ────────────────────────────────────────────\n  // Restrict which MCP tools the model can call. If omitted, all\n  // tools discovered from the server are available.\n  // capabilities: [\"tool_a\", \"tool_b\"],\n\n  prompt: {\n    text: `Use the available MCP tools to complete the task.\nReplace this prompt with instructions specific to your MCP server.`,\n  },\n\n  assertions: [\n    {\n      type: \"llm-rubric\",\n      template: \"mcp-output-correctness\",\n      criteria: [\n        \"Correctly discovers and selects the appropriate tool\",\n        \"Passes valid arguments to the tool\",\n        \"Interprets the tool response coherently\",\n      ],\n    },\n  ],\n\n  status: \"draft\",\n})\n";
+/** Generated YAML for example-mcp-tool-usage (from parsed TS data) */
+export const exampleMcpToolUsageYaml = "- mode: mcp-server\n  id: example-mcp-tool-usage\n  title: MCP tool discovery and invocation\n  description: Example — tests MCP server tool-use (draft)\n  area: mcp\n  serverConfig:\n    transport: streamable-http\n    url: https://your-mcp-server.example.com\n    headers:\n      Authorization: Bearer {{env.MCP_AUTH_TOKEN}}\n  prompt:\n    text: |-\n      Use the available MCP tools to complete the task.\n      Replace this prompt with instructions specific to your MCP server.\n  assertions:\n    - type: llm-rubric\n      template: mcp-output-correctness\n      criteria:\n        - Correctly discovers and selects the appropriate tool\n        - Passes valid arguments to the tool\n        - Interprets the tool response coherently\n  status: draft\n";
+/** Parsed task data for example-path-based-ref (JSON-safe) */
+export const examplePathBasedRefData = [
+    {
+        "mode": "literacy",
+        "id": "example-path-based-ref",
+        "title": "GROQ mutations (path-based doc references)",
+        "description": "Example — demonstrates path-based canonical doc references",
+        "area": "groq",
+        "context": {
+            "docs": [
+                {
+                    "path": "content-lake/mutations-introduction",
+                    "reason": "Introduction to document mutations in the Content Lake"
+                },
+                {
+                    "path": "content-lake/documents",
+                    "reason": "Document structure and types (Content Lake, not CLI reference)"
+                }
+            ]
+        },
         "docCoverage": true,
-        "vars": {
-            "task": "Explain how to create, update, and delete documents in Sanity's\nContent Lake using mutations. Cover:\n1. The different mutation types (create, createOrReplace, patch, delete)\n2. Document structure and required fields (_id, _type)\n3. How to use patch operations to update specific fields\n4. Best practices for mutation patterns\nProvide working code examples using @sanity/client.\n",
-            "docs": ""
+        "prompt": {
+            "text": "Explain how to create, update, and delete documents in Sanity's\nContent Lake using mutations. Cover:\n1. The different mutation types (create, createOrReplace, patch, delete)\n2. Document structure and required fields (_id, _type)\n3. How to use patch operations to update specific fields\n4. Best practices for mutation patterns\nProvide working code examples using @sanity/client."
         },
-        "assert": [
+        "assertions": [
             {
                 "type": "llm-rubric",
                 "template": "task-completion",
@@ -291,30 +370,35 @@ export const examplePathBasedRefData = [
         "status": "draft"
     }
 ];
-/** Raw YAML string for example-path-based-ref (preserves comments) */
-export const examplePathBasedRefYaml = "# ──────────────────────────────────────────────────────────────────────\n# Example Task: Path-based canonical doc references\n# ──────────────────────────────────────────────────────────────────────\n#\n# Demonstrates using `path` to reference canonical documentation.\n# Paths are the preferred reference type because they uniquely identify\n# an article across sections (unlike slugs, which can collide).\n#\n# Path format:\n#   - Simple:   \"webhooks\"              → resolves by slug lookup\n#   - Sectioned: \"content-lake/webhooks\" → disambiguates by section + slug\n#\n# This example demonstrates why paths matter: the slug \"documents\"\n# exists in both the \"content-lake\" and \"cli-reference\" sections.\n# Using \"content-lake/documents\" ensures we get the right one.\n#\n# This example task ships as a DRAFT so it does not run in production\n#   evaluations automatically. To activate it, change status to \"active\"\n# or remove the status line entirely (defaults to active).\n#\n# @see docs/design-docs/canonical-doc-resolution.md\n# ──────────────────────────────────────────────────────────────────────\n\n- id: example-path-based-ref\n  description: \"Example — GROQ mutations (path-based doc references)\"\n\n  featureArea: groq\n\n  # Path-based canonical doc references.\n  #\n  # Use \"section/slug\" format to uniquely identify articles:\n  #   - \"content-lake/mutations-introduction\" → the mutations article\n  #   - \"content-lake/documents\" → the documents article in Content Lake\n  #     (not the CLI \"documents\" article in cli-reference section)\n  #\n  # The \"documents\" slug exists in two sections — this is exactly why\n  # path-based references are preferred over slug-based references.\n  canonicalDocs:\n    - path: content-lake/mutations-introduction\n      reason: \"Introduction to document mutations in the Content Lake\"\n    - path: content-lake/documents\n      reason: \"Document structure and types (Content Lake, not CLI reference)\"\n\n  docCoverage: true\n\n  vars:\n    task: |\n      Explain how to create, update, and delete documents in Sanity's\n      Content Lake using mutations. Cover:\n      1. The different mutation types (create, createOrReplace, patch, delete)\n      2. Document structure and required fields (_id, _type)\n      3. How to use patch operations to update specific fields\n      4. Best practices for mutation patterns\n      Provide working code examples using @sanity/client.\n    docs: \"\"\n\n  assert:\n    - type: llm-rubric\n      template: task-completion\n      criteria:\n        - \"Explains create, createOrReplace, patch, and delete mutations\"\n        - \"Describes required document fields (_id, _type)\"\n        - \"Shows patch operations for field-level updates\"\n        - \"Includes practical code examples\"\n\n    - type: llm-rubric\n      template: code-correctness\n      criteria:\n        - \"Uses correct @sanity/client mutation API\"\n        - \"Patch operations use valid set/unset/inc syntax\"\n\n  baseline:\n    enabled: true\n    rubric: full\n\n  # Example tasks ship as drafts so they don't run in production evals.\n  # Change to \"active\" (or remove this line) to activate.\n  status: draft\n";
+/** TypeScript task template for example-path-based-ref */
+export const examplePathBasedRefTs = "/**\n * Example Task: Path-based canonical doc references.\n *\n * Demonstrates using `path` to reference canonical documentation.\n * Paths are the preferred reference type because they uniquely identify\n * an article across sections (unlike slugs, which can collide).\n *\n * Path format:\n *   - Simple:    \"webhooks\"                → resolves by slug lookup\n *   - Sectioned: \"content-lake/webhooks\"   → disambiguates by section + slug\n *\n * This example demonstrates why paths matter: the slug \"documents\"\n * exists in both the \"content-lake\" and \"cli-reference\" sections.\n * Using \"content-lake/documents\" ensures we get the right one.\n *\n * This example task ships as a DRAFT so it does not run in production\n * evaluations automatically. To activate it, change status to \"active\"\n * or remove the status field entirely (defaults to active).\n *\n * @see docs/design-docs/canonical-doc-resolution.md\n */\n\nimport { defineTask } from \"@sanity/ailf-core\"\n\nexport default defineTask({\n  mode: \"literacy\",\n  id: \"example-path-based-ref\",\n  title: \"GROQ mutations (path-based doc references)\",\n  description: \"Example — demonstrates path-based canonical doc references\",\n\n  area: \"groq\",\n\n  // Path-based canonical doc references.\n  //\n  // Use \"section/slug\" format to uniquely identify articles:\n  //   - \"content-lake/mutations-introduction\" → the mutations article\n  //   - \"content-lake/documents\" → the documents article in Content Lake\n  //     (not the CLI \"documents\" article in cli-reference section)\n  //\n  // The \"documents\" slug exists in two sections — this is exactly why\n  // path-based references are preferred over slug-based references.\n  context: {\n    docs: [\n      {\n        path: \"content-lake/mutations-introduction\",\n        reason: \"Introduction to document mutations in the Content Lake\",\n      },\n      {\n        path: \"content-lake/documents\",\n        reason:\n          \"Document structure and types (Content Lake, not CLI reference)\",\n      },\n    ],\n  },\n\n  docCoverage: true,\n\n  prompt: {\n    text: `Explain how to create, update, and delete documents in Sanity's\nContent Lake using mutations. Cover:\n1. The different mutation types (create, createOrReplace, patch, delete)\n2. Document structure and required fields (_id, _type)\n3. How to use patch operations to update specific fields\n4. Best practices for mutation patterns\nProvide working code examples using @sanity/client.`,\n  },\n\n  assertions: [\n    {\n      type: \"llm-rubric\",\n      template: \"task-completion\",\n      criteria: [\n        \"Explains create, createOrReplace, patch, and delete mutations\",\n        \"Describes required document fields (_id, _type)\",\n        \"Shows patch operations for field-level updates\",\n        \"Includes practical code examples\",\n      ],\n    },\n    {\n      type: \"llm-rubric\",\n      template: \"code-correctness\",\n      criteria: [\n        \"Uses correct @sanity/client mutation API\",\n        \"Patch operations use valid set/unset/inc syntax\",\n      ],\n    },\n  ],\n\n  baseline: { enabled: true, rubric: \"full\" },\n  status: \"draft\",\n})\n";
+/** Generated YAML for example-path-based-ref (from parsed TS data) */
+export const examplePathBasedRefYaml = "- mode: literacy\n  id: example-path-based-ref\n  title: GROQ mutations (path-based doc references)\n  description: Example — demonstrates path-based canonical doc references\n  area: groq\n  context:\n    docs:\n      - path: content-lake/mutations-introduction\n        reason: Introduction to document mutations in the Content Lake\n      - path: content-lake/documents\n        reason: Document structure and types (Content Lake, not CLI reference)\n  docCoverage: true\n  prompt:\n    text: |-\n      Explain how to create, update, and delete documents in Sanity's\n      Content Lake using mutations. Cover:\n      1. The different mutation types (create, createOrReplace, patch, delete)\n      2. Document structure and required fields (_id, _type)\n      3. How to use patch operations to update specific fields\n      4. Best practices for mutation patterns\n      Provide working code examples using @sanity/client.\n  assertions:\n    - type: llm-rubric\n      template: task-completion\n      criteria:\n        - Explains create, createOrReplace, patch, and delete mutations\n        - Describes required document fields (_id, _type)\n        - Shows patch operations for field-level updates\n        - Includes practical code examples\n    - type: llm-rubric\n      template: code-correctness\n      criteria:\n        - Uses correct @sanity/client mutation API\n        - Patch operations use valid set/unset/inc syntax\n  baseline:\n    enabled: true\n    rubric: full\n  status: draft\n";
 /** Parsed task data for example-perspective-ref (JSON-safe) */
 export const examplePerspectiveRefData = [
     {
+        "mode": "literacy",
         "id": "example-perspective-ref",
-        "description": "Example — GROQ features from content release (perspective-based doc references)",
-        "featureArea": "groq",
-        "canonicalDocs": [
-            {
-                "perspective": "rE9TSJvR4",
-                "reason": "All GROQ documentation updates in the test content release"
-            },
-            {
-                "slug": "groq-data-types",
-                "reason": "GROQ data type reference (published, stable)"
-            }
-        ],
+        "title": "GROQ features from content release (perspective-based doc references)",
+        "description": "Example — demonstrates perspective-based canonical doc references",
+        "area": "groq",
+        "context": {
+            "docs": [
+                {
+                    "perspective": "rE9TSJvR4",
+                    "reason": "All GROQ documentation updates in the test content release"
+                },
+                {
+                    "slug": "groq-data-types",
+                    "reason": "GROQ data type reference (published, stable)"
+                }
+            ]
+        },
         "docCoverage": true,
-        "vars": {
-            "task": "Using GROQ, demonstrate advanced query patterns including:\n1. Joining data across document types using references\n2. Filtering webhook payloads with GROQ projections\n3. Using the query cheat sheet patterns for common operations\n4. Working with different GROQ data types in filters\nProvide working GROQ query examples for each pattern.\n",
-            "docs": ""
+        "prompt": {
+            "text": "Using GROQ, demonstrate advanced query patterns including:\n1. Joining data across document types using references\n2. Filtering webhook payloads with GROQ projections\n3. Using the query cheat sheet patterns for common operations\n4. Working with different GROQ data types in filters\nProvide working GROQ query examples for each pattern."
         },
-        "assert": [
+        "assertions": [
             {
                 "type": "llm-rubric",
                 "template": "task-completion",
@@ -340,31 +424,36 @@ export const examplePerspectiveRefData = [
         "status": "draft"
     }
 ];
-/** Raw YAML string for example-perspective-ref (preserves comments) */
-export const examplePerspectiveRefYaml = "# ──────────────────────────────────────────────────────────────────────\n# Example Task: Perspective / content release doc references\n# ──────────────────────────────────────────────────────────────────────\n#\n# Demonstrates using `perspective` to reference all documentation\n# articles within a content release. This is the key capability for\n# evaluating NEW feature documentation before it's published.\n#\n# How it works:\n#   - A perspective ref is one-to-many: the doc fetcher queries the\n#     named release and expands it to ALL articles versioned within it.\n#   - Downstream consumers see the same flat DocContext[] regardless\n#     of how docs were resolved.\n#   - When the release is published, the perspective entry becomes a\n#     no-op (articles are now in published). Migrate to explicit path\n#     or slug refs at your convenience.\n#\n# This example task ships as a DRAFT so it does not run in production\n#   evaluations automatically. To activate it, change status to \"active\"\n# or remove the status line entirely (defaults to active).\n#\n# @see docs/design-docs/canonical-doc-resolution.md\n# ──────────────────────────────────────────────────────────────────────\n\n- id: example-perspective-ref\n  description:\n    \"Example — GROQ features from content release (perspective-based doc\n    references)\"\n\n  featureArea: groq\n\n  # Perspective-based canonical doc reference.\n  #\n  # The perspective ID references a content release in the Sanity\n  # Content Lake. At evaluation time, the doc fetcher auto-discovers\n  # all articles versioned in this release and includes them as\n  # canonical documentation context.\n  #\n  # Release rE9TSJvR4 contains:\n  #   - \"GROQ-powered webhooks\" (webhooks)\n  #   - \"Query Cheat Sheet - GROQ\" (query-cheat-sheet)\n  #   - \"GROQ joins\" (groq-joins)\n  #\n  # You can combine perspective refs with explicit slug/path/id refs\n  # to include foundational published docs alongside release content.\n  # Here we add groq-data-types as a complementary published reference.\n  canonicalDocs:\n    - perspective: rE9TSJvR4\n      reason: \"All GROQ documentation updates in the test content release\"\n    - slug: groq-data-types\n      reason: \"GROQ data type reference (published, stable)\"\n\n  docCoverage: true\n\n  vars:\n    task: |\n      Using GROQ, demonstrate advanced query patterns including:\n      1. Joining data across document types using references\n      2. Filtering webhook payloads with GROQ projections\n      3. Using the query cheat sheet patterns for common operations\n      4. Working with different GROQ data types in filters\n      Provide working GROQ query examples for each pattern.\n    docs: \"\"\n\n  assert:\n    - type: llm-rubric\n      template: task-completion\n      criteria:\n        - \"Demonstrates GROQ join syntax for cross-document queries\"\n        - \"Shows GROQ filter patterns for webhook configuration\"\n        - \"Includes practical query examples from cheat sheet patterns\"\n\n    - type: llm-rubric\n      template: code-correctness\n      criteria:\n        - \"All GROQ queries use valid syntax\"\n        - \"Reference joins use correct dereference operator (->)\"\n\n  baseline:\n    enabled: true\n    rubric: full\n\n  # Example tasks ship as drafts so they don't run in production evals.\n  # Change to \"active\" (or remove this line) to activate.\n  status: draft\n";
+/** TypeScript task template for example-perspective-ref */
+export const examplePerspectiveRefTs = "/**\n * Example Task: Perspective / content release doc references.\n *\n * Demonstrates using `perspective` to reference all documentation\n * articles within a content release. This is the key capability for\n * evaluating NEW feature documentation before it's published.\n *\n * How it works:\n *   - A perspective ref is one-to-many: the doc fetcher queries the\n *     named release and expands it to ALL articles versioned within it.\n *   - Downstream consumers see the same flat DocContext[] regardless\n *     of how docs were resolved.\n *   - When the release is published, the perspective entry becomes a\n *     no-op (articles are now in published). Migrate to explicit path\n *     or slug refs at your convenience.\n *\n * This example task ships as a DRAFT so it does not run in production\n * evaluations automatically. To activate it, change status to \"active\"\n * or remove the status field entirely (defaults to active).\n *\n * @see docs/design-docs/canonical-doc-resolution.md\n */\n\nimport { defineTask } from \"@sanity/ailf-core\"\n\nexport default defineTask({\n  mode: \"literacy\",\n  id: \"example-perspective-ref\",\n  title:\n    \"GROQ features from content release (perspective-based doc references)\",\n  description:\n    \"Example — demonstrates perspective-based canonical doc references\",\n\n  area: \"groq\",\n\n  // Perspective-based canonical doc reference.\n  //\n  // The perspective ID references a content release in the Sanity\n  // Content Lake. At evaluation time, the doc fetcher auto-discovers\n  // all articles versioned in this release and includes them as\n  // canonical documentation context.\n  //\n  // Release rE9TSJvR4 contains:\n  //   - \"GROQ-powered webhooks\" (webhooks)\n  //   - \"Query Cheat Sheet - GROQ\" (query-cheat-sheet)\n  //   - \"GROQ joins\" (groq-joins)\n  //\n  // You can combine perspective refs with explicit slug/path/id refs\n  // to include foundational published docs alongside release content.\n  // Here we add groq-data-types as a complementary published reference.\n  context: {\n    docs: [\n      {\n        perspective: \"rE9TSJvR4\",\n        reason: \"All GROQ documentation updates in the test content release\",\n      },\n      {\n        slug: \"groq-data-types\",\n        reason: \"GROQ data type reference (published, stable)\",\n      },\n    ],\n  },\n\n  docCoverage: true,\n\n  prompt: {\n    text: `Using GROQ, demonstrate advanced query patterns including:\n1. Joining data across document types using references\n2. Filtering webhook payloads with GROQ projections\n3. Using the query cheat sheet patterns for common operations\n4. Working with different GROQ data types in filters\nProvide working GROQ query examples for each pattern.`,\n  },\n\n  assertions: [\n    {\n      type: \"llm-rubric\",\n      template: \"task-completion\",\n      criteria: [\n        \"Demonstrates GROQ join syntax for cross-document queries\",\n        \"Shows GROQ filter patterns for webhook configuration\",\n        \"Includes practical query examples from cheat sheet patterns\",\n      ],\n    },\n    {\n      type: \"llm-rubric\",\n      template: \"code-correctness\",\n      criteria: [\n        \"All GROQ queries use valid syntax\",\n        \"Reference joins use correct dereference operator (->)\",\n      ],\n    },\n  ],\n\n  baseline: { enabled: true, rubric: \"full\" },\n  status: \"draft\",\n})\n";
+/** Generated YAML for example-perspective-ref (from parsed TS data) */
+export const examplePerspectiveRefYaml = "- mode: literacy\n  id: example-perspective-ref\n  title: GROQ features from content release (perspective-based doc references)\n  description: Example — demonstrates perspective-based canonical doc references\n  area: groq\n  context:\n    docs:\n      - perspective: rE9TSJvR4\n        reason: All GROQ documentation updates in the test content release\n      - slug: groq-data-types\n        reason: GROQ data type reference (published, stable)\n  docCoverage: true\n  prompt:\n    text: |-\n      Using GROQ, demonstrate advanced query patterns including:\n      1. Joining data across document types using references\n      2. Filtering webhook payloads with GROQ projections\n      3. Using the query cheat sheet patterns for common operations\n      4. Working with different GROQ data types in filters\n      Provide working GROQ query examples for each pattern.\n  assertions:\n    - type: llm-rubric\n      template: task-completion\n      criteria:\n        - Demonstrates GROQ join syntax for cross-document queries\n        - Shows GROQ filter patterns for webhook configuration\n        - Includes practical query examples from cheat sheet patterns\n    - type: llm-rubric\n      template: code-correctness\n      criteria:\n        - All GROQ queries use valid syntax\n        - Reference joins use correct dereference operator (->)\n  baseline:\n    enabled: true\n    rubric: full\n  status: draft\n";
 /** Parsed task data for example-studio-custom-input (JSON-safe) */
 export const exampleStudioCustomInputData = [
     {
+        "mode": "literacy",
         "id": "example-studio-custom-input",
-        "description": "Example — Custom input component in Sanity Studio",
-        "featureArea": "studio",
-        "canonicalDocs": [
-            {
-                "slug": "custom-input-widgets",
-                "reason": "Guide for building custom form inputs in Sanity Studio"
-            },
-            {
-                "slug": "form-components",
-                "reason": "Form component API and customization patterns"
-            }
-        ],
+        "title": "Custom input component in Sanity Studio",
+        "description": "Example — tests Studio custom input implementation",
+        "area": "studio",
+        "context": {
+            "docs": [
+                {
+                    "slug": "custom-input-widgets",
+                    "reason": "Guide for building custom form inputs in Sanity Studio"
+                },
+                {
+                    "slug": "form-components",
+                    "reason": "Form component API and customization patterns"
+                }
+            ]
+        },
         "docCoverage": true,
         "referenceSolution": "canonical/example-studio-custom-input.ts",
-        "vars": {
-            "task": "Build a custom string input component for Sanity Studio that shows\na character count below the input field. The component should accept\na maxLength option from the field schema and display a warning when\nthe text exceeds the limit.\n",
-            "docs": ""
+        "prompt": {
+            "text": "Build a custom string input component for Sanity Studio that shows\na character count below the input field. The component should accept\na maxLength option from the field schema and display a warning when\nthe text exceeds the limit."
         },
-        "assert": [
+        "assertions": [
             {
                 "type": "llm-rubric",
                 "template": "task-completion",
@@ -391,8 +480,10 @@ export const exampleStudioCustomInputData = [
         "status": "draft"
     }
 ];
-/** Raw YAML string for example-studio-custom-input (preserves comments) */
-export const exampleStudioCustomInputYaml = "# ──────────────────────────────────────────────────────────────────────\n# Example Task: Custom input component in Sanity Studio\n# ──────────────────────────────────────────────────────────────────────\n#\n# This is a starter template — edit it for your own documentation.\n# Delete this file or replace it with your own tasks.\n#\n# This example task ships as a DRAFT so it does not run in production\n#   evaluations automatically. To activate it, change status to \"active\"\n# or remove the status line entirely (defaults to active).\n# ──────────────────────────────────────────────────────────────────────\n\n- id: example-studio-custom-input\n  description: \"Example — Custom input component in Sanity Studio\"\n\n  featureArea: studio\n\n  # Slug-based canonical doc references.\n  canonicalDocs:\n    - slug: custom-input-widgets\n      reason: \"Guide for building custom form inputs in Sanity Studio\"\n    - slug: form-components\n      reason: \"Form component API and customization patterns\"\n\n  docCoverage: true\n  referenceSolution: canonical/example-studio-custom-input.ts\n\n  vars:\n    task: |\n      Build a custom string input component for Sanity Studio that shows\n      a character count below the input field. The component should accept\n      a maxLength option from the field schema and display a warning when\n      the text exceeds the limit.\n    docs: \"\"\n\n  assert:\n    - type: llm-rubric\n      template: task-completion\n      criteria:\n        - \"Implements a React component that renders a text input\"\n        - \"Displays a live character count\"\n        - \"Reads maxLength from schema options\"\n        - \"Shows a visual warning when limit is exceeded\"\n\n    - type: llm-rubric\n      template: code-correctness\n      criteria:\n        - \"Uses the Sanity UI library for styling\"\n        - \"Calls onChange with patch operations\"\n\n  baseline:\n    enabled: true\n    rubric: full\n\n  # Example tasks ship as drafts so they don't run in production evals.\n  # Change to \"active\" (or remove this line) to activate.\n  status: draft\n";
+/** TypeScript task template for example-studio-custom-input */
+export const exampleStudioCustomInputTs = "/**\n * Example Task: Custom input component in Sanity Studio.\n *\n * This is a starter template — edit it for your own documentation.\n * Delete this file or replace it with your own tasks.\n *\n * This example task ships as a DRAFT so it does not run in production\n * evaluations automatically. To activate it, change status to \"active\"\n * or remove the status field entirely (defaults to active).\n */\n\nimport { defineTask } from \"@sanity/ailf-core\"\n\nexport default defineTask({\n  mode: \"literacy\",\n  id: \"example-studio-custom-input\",\n  title: \"Custom input component in Sanity Studio\",\n  description: \"Example — tests Studio custom input implementation\",\n\n  area: \"studio\",\n\n  context: {\n    docs: [\n      {\n        slug: \"custom-input-widgets\",\n        reason: \"Guide for building custom form inputs in Sanity Studio\",\n      },\n      {\n        slug: \"form-components\",\n        reason: \"Form component API and customization patterns\",\n      },\n    ],\n  },\n\n  docCoverage: true,\n  referenceSolution: \"canonical/example-studio-custom-input.ts\",\n\n  prompt: {\n    text: `Build a custom string input component for Sanity Studio that shows\na character count below the input field. The component should accept\na maxLength option from the field schema and display a warning when\nthe text exceeds the limit.`,\n  },\n\n  assertions: [\n    {\n      type: \"llm-rubric\",\n      template: \"task-completion\",\n      criteria: [\n        \"Implements a React component that renders a text input\",\n        \"Displays a live character count\",\n        \"Reads maxLength from schema options\",\n        \"Shows a visual warning when limit is exceeded\",\n      ],\n    },\n    {\n      type: \"llm-rubric\",\n      template: \"code-correctness\",\n      criteria: [\n        \"Uses the Sanity UI library for styling\",\n        \"Calls onChange with patch operations\",\n      ],\n    },\n  ],\n\n  baseline: { enabled: true, rubric: \"full\" },\n  status: \"draft\",\n})\n";
+/** Generated YAML for example-studio-custom-input (from parsed TS data) */
+export const exampleStudioCustomInputYaml = "- mode: literacy\n  id: example-studio-custom-input\n  title: Custom input component in Sanity Studio\n  description: Example — tests Studio custom input implementation\n  area: studio\n  context:\n    docs:\n      - slug: custom-input-widgets\n        reason: Guide for building custom form inputs in Sanity Studio\n      - slug: form-components\n        reason: Form component API and customization patterns\n  docCoverage: true\n  referenceSolution: canonical/example-studio-custom-input.ts\n  prompt:\n    text: |-\n      Build a custom string input component for Sanity Studio that shows\n      a character count below the input field. The component should accept\n      a maxLength option from the field schema and display a warning when\n      the text exceeds the limit.\n  assertions:\n    - type: llm-rubric\n      template: task-completion\n      criteria:\n        - Implements a React component that renders a text input\n        - Displays a live character count\n        - Reads maxLength from schema options\n        - Shows a visual warning when limit is exceeded\n    - type: llm-rubric\n      template: code-correctness\n      criteria:\n        - Uses the Sanity UI library for styling\n        - Calls onChange with patch operations\n  baseline:\n    enabled: true\n    rubric: full\n  status: draft\n";
 // ---------------------------------------------------------------------------
 // Aggregate task exports
 // ---------------------------------------------------------------------------
@@ -400,20 +491,43 @@ export const exampleStudioCustomInputYaml = "# ───────────
 export const allTaskData = [
     ...exampleGroqBlogListingData,
     ...exampleIdBasedRefData,
+    ...exampleKnowledgeProbeData,
+    ...exampleMcpToolUsageData,
     ...examplePathBasedRefData,
     ...examplePerspectiveRefData,
     ...exampleStudioCustomInputData,
 ];
-/** Map of task ID (filename stem) → raw YAML string (preserves comments) */
+/** Map of task ID (filename stem) → raw TypeScript source */
+export const taskTsFiles = {
+    "example-groq-blog-listing": exampleGroqBlogListingTs,
+    "example-id-based-ref": exampleIdBasedRefTs,
+    "example-knowledge-probe": exampleKnowledgeProbeTs,
+    "example-mcp-tool-usage": exampleMcpToolUsageTs,
+    "example-path-based-ref": examplePathBasedRefTs,
+    "example-perspective-ref": examplePerspectiveRefTs,
+    "example-studio-custom-input": exampleStudioCustomInputTs,
+};
+/** Map of task ID (filename stem) → generated YAML string */
 export const taskYamlFiles = {
     "example-groq-blog-listing": exampleGroqBlogListingYaml,
     "example-id-based-ref": exampleIdBasedRefYaml,
+    "example-knowledge-probe": exampleKnowledgeProbeYaml,
+    "example-mcp-tool-usage": exampleMcpToolUsageYaml,
     "example-path-based-ref": examplePathBasedRefYaml,
     "example-perspective-ref": examplePerspectiveRefYaml,
     "example-studio-custom-input": exampleStudioCustomInputYaml,
 };
 /** List of task file stems, in alphabetical order */
-export const TASK_FILE_NAMES = ["example-groq-blog-listing", "example-id-based-ref", "example-path-based-ref", "example-perspective-ref", "example-studio-custom-input"];
+export const TASK_FILE_NAMES = ["example-groq-blog-listing", "example-id-based-ref", "example-knowledge-probe", "example-mcp-tool-usage", "example-path-based-ref", "example-perspective-ref", "example-studio-custom-input"];
+export const TASK_EXAMPLES = [
+    { stem: "example-groq-blog-listing", mode: "literacy", status: "draft" },
+    { stem: "example-id-based-ref", mode: "literacy", status: "draft" },
+    { stem: "example-knowledge-probe", mode: "knowledge-probe", status: "draft" },
+    { stem: "example-mcp-tool-usage", mode: "mcp-server", status: "draft" },
+    { stem: "example-path-based-ref", mode: "literacy", status: "draft" },
+    { stem: "example-perspective-ref", mode: "literacy", status: "draft" },
+    { stem: "example-studio-custom-input", mode: "literacy", status: "draft" },
+];
 export const EXAMPLE_TYPES = ["config", "source", "rubric", "threshold", "ailf-config", "task"];
 export const EXAMPLES = {
     "config": {
@@ -457,23 +571,3 @@ export const workflowYaml = "# ────────────────
 // ---------------------------------------------------------------------------
 /** TypeScript project configuration template (ailf.config.ts) */
 export const ailfConfigTs = "/**\n * .ailf/ailf.config.ts — AI Literacy Framework project configuration.\n *\n * This file configures how the AILF evaluation pipeline runs in this\n * repository. Place it at .ailf/ailf.config.ts in your project root.\n *\n * Evaluations are submitted to the AILF API (ailf-api.sanity.build).\n * The API handles LLM calls, doc fetching, grading, and report\n * publishing. Your repo only needs one secret: AILF_API_KEY.\n *\n * Docs: https://github.com/sanity-labs/ai-literacy-framework\n */\n\nexport default {\n  /**\n   * Documentation source — which docs are being evaluated.\n   *\n   * This tells the pipeline which Sanity project and dataset contain\n   * the documentation under test. For most users, this is Sanity's own\n   * docs project.\n   */\n  source: {\n    /** Sanity project ID (find yours at sanity.io/manage) */\n    projectId: \"3do82whm\",\n    /** The dataset to query (e.g., \"production\", \"next\") */\n    dataset: \"next\",\n    /**\n     * The public URL of your documentation site.\n     * Used by agentic mode to test agent discoverability.\n     */\n    baseUrl: \"https://www.sanity.io/docs\",\n  },\n\n  /**\n   * Trigger configuration — when evaluations run automatically.\n   *\n   * Each key is a trigger context. The pipeline checks which trigger\n   * matches the current execution context (PR, merge, schedule, etc.)\n   * and applies its settings.\n   *\n   * Mode options:\n   *   \"validate-only\" — check that task files parse correctly (fast, no LLM calls)\n   *   \"eval\"          — run the full evaluation pipeline\n   */\n  triggers: {\n    /** On pull requests: just validate task files parse correctly. */\n    pr: {\n      mode: \"validate-only\",\n    },\n\n    /** When .ailf/ files change in a PR: run a real evaluation. */\n    \"pr-task-change\": {\n      mode: \"eval\",\n      paths: [\".ailf/**\"],\n    },\n\n    /** On merge to main: run evaluation (non-blocking). */\n    main: {\n      mode: \"eval\",\n      blocking: false,\n      notify: true,\n    },\n  },\n}\n";
-/** TypeScript task template for example-groq-blog-listing */
-export const exampleGroqBlogListingTs = "/**\n * Example Task: Blog listing with GROQ queries.\n *\n * This is a starter template — edit it for your own documentation.\n * Each task evaluates whether an AI coding agent can implement a feature\n * using your docs as context. Delete this file or replace it entirely.\n *\n * This example task ships as a DRAFT so it does not run in production\n * evaluations automatically. To activate it, change status to \"active\"\n * or remove the status field entirely (defaults to active).\n *\n * Full field reference:\n *   https://github.com/sanity-labs/ai-literacy-framework/blob/main/docs/CONTRIBUTING_TASKS.md\n */\n\nimport { defineTask } from \"@sanity/ailf-core\"\n\nexport default defineTask({\n  // ── Mode ────────────────────────────────────────────────────────────\n  // \"literacy\" tests whether AI coding tools can implement features\n  // using your docs as context. Other modes: \"mcp-server\",\n  // \"knowledge-probe\", \"agent-harness\", \"custom\".\n  mode: \"literacy\",\n\n  // ── Identity ────────────────────────────────────────────────────────\n  // Unique identifier — lowercase alphanumeric with hyphens.\n  // Must be unique across all task files in .ailf/tasks/.\n  id: \"example-groq-blog-listing\",\n  title: \"Blog listing with GROQ queries\",\n  description: \"Example — tests GROQ blog listing implementation\",\n\n  // Feature area this task belongs to. Tasks with the same area are\n  // grouped together in score summaries.\n  area: \"groq\",\n\n  // ── Documentation context ───────────────────────────────────────────\n  // Canonical doc references for this task. The pipeline fetches these\n  // from Sanity and injects them into the prompt for baseline evaluation.\n  //\n  // This example uses slug-based references — the simplest form.\n  // See the other example tasks for path, id, and perspective references.\n  context: {\n    docs: [\n      {\n        slug: \"groq-introduction\",\n        reason: \"Core GROQ syntax and query language reference\",\n      },\n      {\n        slug: \"how-queries-work\",\n        reason: \"Query execution model and best practices\",\n      },\n    ],\n  },\n\n  // When true, the pipeline auto-generates an additional rubric that\n  // checks whether the LLM's response actually used the provided docs.\n  docCoverage: true,\n\n  // Path to a gold-standard implementation, relative to canonical/.\n  // The grader uses this as a reference when scoring code correctness.\n  referenceSolution: \"canonical/example-groq-blog-listing.ts\",\n\n  // ── Prompt ──────────────────────────────────────────────────────────\n  // prompt.text — the implementation prompt given to the LLM.\n  // Write this as if you're asking a developer to build the feature.\n  // Be specific about requirements so the grader can evaluate clearly.\n  prompt: {\n    text: `Create a Next.js page component that lists blog posts from Sanity\nusing GROQ. The page should display the title, slug, and published\ndate for each post, sorted by most recent first. Use the Sanity\nclient to fetch data.`,\n  },\n\n  // ── Assertions ──────────────────────────────────────────────────────\n  // Grading assertions — how the LLM's response is scored.\n  //\n  // \"llm-rubric\" assertions use a grader LLM to score against criteria.\n  // The \"template\" references a rubric template (e.g. task-completion).\n  //\n  // Available templates:\n  //   task-completion   — did the LLM implement the feature? (weight: 0.50)\n  //   code-correctness  — is the code idiomatic and correct? (weight: 0.25)\n  //\n  // You can also use value-based assertions:\n  //   { type: \"contains\", value: \"client.fetch\" }\n  //   { type: \"contains-any\", value: [\"createClient\", \"sanityClient\"] }\n  assertions: [\n    {\n      type: \"llm-rubric\",\n      template: \"task-completion\",\n      criteria: [\n        \"Uses the groq tagged template literal\",\n        \"Fetches blog posts with title, slug, and publishedAt fields\",\n        \"Orders results by publishedAt in descending order\",\n      ],\n    },\n    {\n      type: \"llm-rubric\",\n      template: \"code-correctness\",\n      criteria: [\n        \"Uses createClient from @sanity/client or next-sanity\",\n        \"Exports a valid Next.js page component\",\n      ],\n    },\n  ],\n\n  // ── Baseline variant ────────────────────────────────────────────────\n  //   enabled — set to false to skip this task entirely\n  //   rubric  — \"full\" (default), \"abbreviated\" (faster), or \"none\"\n  baseline: {\n    enabled: true,\n    rubric: \"full\",\n  },\n\n  // Example tasks ship as drafts so they don't run in production evals.\n  // Change to \"active\" (or remove this field) to activate.\n  status: \"draft\",\n})\n";
-/** TypeScript task template for example-id-based-ref */
-export const exampleIdBasedRefTs = "/**\n * Example Task: Document ID-based canonical doc references.\n *\n * Demonstrates using `id` to reference canonical documentation by\n * Sanity document `_id`. This is useful for:\n *   - Draft documents that don't have a stable slug yet\n *   - Programmatic references from imports or migrations\n *   - Documents where you know the _id but not the slug\n *\n * The `id` ref type can also carry optional `slug` and `path` fields\n * as human-readable annotations — these are NOT used for resolution,\n * only for display in logs and reports.\n *\n * This example task ships as a DRAFT so it does not run in production\n * evaluations automatically. To activate it, change status to \"active\"\n * or remove the status field entirely (defaults to active).\n *\n * @see docs/design-docs/canonical-doc-resolution.md\n */\n\nimport { defineTask } from \"@sanity/ailf-core\"\n\nexport default defineTask({\n  mode: \"literacy\",\n  id: \"example-id-based-ref\",\n  title: \"GROQ feature support (ID-based doc references)\",\n  description: \"Example — demonstrates ID-based canonical doc references\",\n\n  area: \"groq\",\n\n  // ID-based canonical doc references.\n  //\n  // Use the Sanity document _id to reference articles directly.\n  // Optional slug/path annotations help humans reading the file\n  // but are NOT used for resolution — only the `id` field matters.\n  //\n  // These IDs reference real articles in the Sanity docs (next dataset):\n  //   0ba88f1b... = \"GROQ feature support across Sanity\"\n  //   5b9c2863... = \"Custom GROQ functions\"\n  context: {\n    docs: [\n      {\n        id: \"0ba88f1b-d1a7-418a-9267-2e343d01886a\",\n        slug: \"groq-feature-support-by-context\", // annotation only\n        reason: \"GROQ feature support across different Sanity contexts\",\n      },\n      {\n        id: \"5b9c2863-ef01-4565-af8e-ee54e081ee74\",\n        slug: \"custom-groq-functions\", // annotation only\n        reason: \"Custom GROQ functions and pipelines\",\n      },\n    ],\n  },\n\n  docCoverage: true,\n\n  prompt: {\n    text: `Explain how GROQ is used across different Sanity contexts.\nCover the following:\n1. Which GROQ features are available in each context (API queries,\n   webhooks, custom functions, access control)\n2. How to create and use custom GROQ functions\n3. Any differences in GROQ support between contexts\nProvide examples demonstrating context-specific GROQ patterns.`,\n  },\n\n  assertions: [\n    {\n      type: \"llm-rubric\",\n      template: \"task-completion\",\n      criteria: [\n        \"Explains GROQ availability across different Sanity contexts\",\n        \"Describes custom GROQ function creation and usage\",\n        \"Notes differences in GROQ support between contexts\",\n      ],\n    },\n    {\n      type: \"llm-rubric\",\n      template: \"code-correctness\",\n      criteria: [\n        \"GROQ examples use valid syntax\",\n        \"Custom function examples follow the correct API pattern\",\n      ],\n    },\n  ],\n\n  baseline: { enabled: true, rubric: \"full\" },\n  status: \"draft\",\n})\n";
-/** TypeScript task template for example-path-based-ref */
-export const examplePathBasedRefTs = "/**\n * Example Task: Path-based canonical doc references.\n *\n * Demonstrates using `path` to reference canonical documentation.\n * Paths are the preferred reference type because they uniquely identify\n * an article across sections (unlike slugs, which can collide).\n *\n * Path format:\n *   - Simple:    \"webhooks\"                → resolves by slug lookup\n *   - Sectioned: \"content-lake/webhooks\"   → disambiguates by section + slug\n *\n * This example demonstrates why paths matter: the slug \"documents\"\n * exists in both the \"content-lake\" and \"cli-reference\" sections.\n * Using \"content-lake/documents\" ensures we get the right one.\n *\n * This example task ships as a DRAFT so it does not run in production\n * evaluations automatically. To activate it, change status to \"active\"\n * or remove the status field entirely (defaults to active).\n *\n * @see docs/design-docs/canonical-doc-resolution.md\n */\n\nimport { defineTask } from \"@sanity/ailf-core\"\n\nexport default defineTask({\n  mode: \"literacy\",\n  id: \"example-path-based-ref\",\n  title: \"GROQ mutations (path-based doc references)\",\n  description: \"Example — demonstrates path-based canonical doc references\",\n\n  area: \"groq\",\n\n  // Path-based canonical doc references.\n  //\n  // Use \"section/slug\" format to uniquely identify articles:\n  //   - \"content-lake/mutations-introduction\" → the mutations article\n  //   - \"content-lake/documents\" → the documents article in Content Lake\n  //     (not the CLI \"documents\" article in cli-reference section)\n  //\n  // The \"documents\" slug exists in two sections — this is exactly why\n  // path-based references are preferred over slug-based references.\n  context: {\n    docs: [\n      {\n        path: \"content-lake/mutations-introduction\",\n        reason: \"Introduction to document mutations in the Content Lake\",\n      },\n      {\n        path: \"content-lake/documents\",\n        reason:\n          \"Document structure and types (Content Lake, not CLI reference)\",\n      },\n    ],\n  },\n\n  docCoverage: true,\n\n  prompt: {\n    text: `Explain how to create, update, and delete documents in Sanity's\nContent Lake using mutations. Cover:\n1. The different mutation types (create, createOrReplace, patch, delete)\n2. Document structure and required fields (_id, _type)\n3. How to use patch operations to update specific fields\n4. Best practices for mutation patterns\nProvide working code examples using @sanity/client.`,\n  },\n\n  assertions: [\n    {\n      type: \"llm-rubric\",\n      template: \"task-completion\",\n      criteria: [\n        \"Explains create, createOrReplace, patch, and delete mutations\",\n        \"Describes required document fields (_id, _type)\",\n        \"Shows patch operations for field-level updates\",\n        \"Includes practical code examples\",\n      ],\n    },\n    {\n      type: \"llm-rubric\",\n      template: \"code-correctness\",\n      criteria: [\n        \"Uses correct @sanity/client mutation API\",\n        \"Patch operations use valid set/unset/inc syntax\",\n      ],\n    },\n  ],\n\n  baseline: { enabled: true, rubric: \"full\" },\n  status: \"draft\",\n})\n";
-/** TypeScript task template for example-perspective-ref */
-export const examplePerspectiveRefTs = "/**\n * Example Task: Perspective / content release doc references.\n *\n * Demonstrates using `perspective` to reference all documentation\n * articles within a content release. This is the key capability for\n * evaluating NEW feature documentation before it's published.\n *\n * How it works:\n *   - A perspective ref is one-to-many: the doc fetcher queries the\n *     named release and expands it to ALL articles versioned within it.\n *   - Downstream consumers see the same flat DocContext[] regardless\n *     of how docs were resolved.\n *   - When the release is published, the perspective entry becomes a\n *     no-op (articles are now in published). Migrate to explicit path\n *     or slug refs at your convenience.\n *\n * This example task ships as a DRAFT so it does not run in production\n * evaluations automatically. To activate it, change status to \"active\"\n * or remove the status field entirely (defaults to active).\n *\n * @see docs/design-docs/canonical-doc-resolution.md\n */\n\nimport { defineTask } from \"@sanity/ailf-core\"\n\nexport default defineTask({\n  mode: \"literacy\",\n  id: \"example-perspective-ref\",\n  title:\n    \"GROQ features from content release (perspective-based doc references)\",\n  description:\n    \"Example — demonstrates perspective-based canonical doc references\",\n\n  area: \"groq\",\n\n  // Perspective-based canonical doc reference.\n  //\n  // The perspective ID references a content release in the Sanity\n  // Content Lake. At evaluation time, the doc fetcher auto-discovers\n  // all articles versioned in this release and includes them as\n  // canonical documentation context.\n  //\n  // Release rE9TSJvR4 contains:\n  //   - \"GROQ-powered webhooks\" (webhooks)\n  //   - \"Query Cheat Sheet - GROQ\" (query-cheat-sheet)\n  //   - \"GROQ joins\" (groq-joins)\n  //\n  // You can combine perspective refs with explicit slug/path/id refs\n  // to include foundational published docs alongside release content.\n  // Here we add groq-data-types as a complementary published reference.\n  context: {\n    docs: [\n      {\n        perspective: \"rE9TSJvR4\",\n        reason: \"All GROQ documentation updates in the test content release\",\n      },\n      {\n        slug: \"groq-data-types\",\n        reason: \"GROQ data type reference (published, stable)\",\n      },\n    ],\n  },\n\n  docCoverage: true,\n\n  prompt: {\n    text: `Using GROQ, demonstrate advanced query patterns including:\n1. Joining data across document types using references\n2. Filtering webhook payloads with GROQ projections\n3. Using the query cheat sheet patterns for common operations\n4. Working with different GROQ data types in filters\nProvide working GROQ query examples for each pattern.`,\n  },\n\n  assertions: [\n    {\n      type: \"llm-rubric\",\n      template: \"task-completion\",\n      criteria: [\n        \"Demonstrates GROQ join syntax for cross-document queries\",\n        \"Shows GROQ filter patterns for webhook configuration\",\n        \"Includes practical query examples from cheat sheet patterns\",\n      ],\n    },\n    {\n      type: \"llm-rubric\",\n      template: \"code-correctness\",\n      criteria: [\n        \"All GROQ queries use valid syntax\",\n        \"Reference joins use correct dereference operator (->)\",\n      ],\n    },\n  ],\n\n  baseline: { enabled: true, rubric: \"full\" },\n  status: \"draft\",\n})\n";
-/** TypeScript task template for example-studio-custom-input */
-export const exampleStudioCustomInputTs = "/**\n * Example Task: Custom input component in Sanity Studio.\n *\n * This is a starter template — edit it for your own documentation.\n * Delete this file or replace it with your own tasks.\n *\n * This example task ships as a DRAFT so it does not run in production\n * evaluations automatically. To activate it, change status to \"active\"\n * or remove the status field entirely (defaults to active).\n */\n\nimport { defineTask } from \"@sanity/ailf-core\"\n\nexport default defineTask({\n  mode: \"literacy\",\n  id: \"example-studio-custom-input\",\n  title: \"Custom input component in Sanity Studio\",\n  description: \"Example — tests Studio custom input implementation\",\n\n  area: \"studio\",\n\n  context: {\n    docs: [\n      {\n        slug: \"custom-input-widgets\",\n        reason: \"Guide for building custom form inputs in Sanity Studio\",\n      },\n      {\n        slug: \"form-components\",\n        reason: \"Form component API and customization patterns\",\n      },\n    ],\n  },\n\n  docCoverage: true,\n  referenceSolution: \"canonical/example-studio-custom-input.ts\",\n\n  prompt: {\n    text: `Build a custom string input component for Sanity Studio that shows\na character count below the input field. The component should accept\na maxLength option from the field schema and display a warning when\nthe text exceeds the limit.`,\n  },\n\n  assertions: [\n    {\n      type: \"llm-rubric\",\n      template: \"task-completion\",\n      criteria: [\n        \"Implements a React component that renders a text input\",\n        \"Displays a live character count\",\n        \"Reads maxLength from schema options\",\n        \"Shows a visual warning when limit is exceeded\",\n      ],\n    },\n    {\n      type: \"llm-rubric\",\n      template: \"code-correctness\",\n      criteria: [\n        \"Uses the Sanity UI library for styling\",\n        \"Calls onChange with patch operations\",\n      ],\n    },\n  ],\n\n  baseline: { enabled: true, rubric: \"full\" },\n  status: \"draft\",\n})\n";
-/** Map of task ID (filename stem) → raw TypeScript source */
-export const taskTsFiles = {
-    "example-groq-blog-listing": exampleGroqBlogListingTs,
-    "example-id-based-ref": exampleIdBasedRefTs,
-    "example-path-based-ref": examplePathBasedRefTs,
-    "example-perspective-ref": examplePerspectiveRefTs,
-    "example-studio-custom-input": exampleStudioCustomInputTs,
-};
-/** List of TS task file stems, in alphabetical order */
-export const TASK_TS_FILE_NAMES = ["example-groq-blog-listing", "example-id-based-ref", "example-path-based-ref", "example-perspective-ref", "example-studio-custom-input"];

package/dist/_vendor/ailf-core/index.d.ts

@@ -18,3 +18,4 @@ export * from "./examples/index.js";
 export { defineConfig, defineFeatures, defineModeBase, defineModels, definePricingTable, definePreset, definePrompts, defineRubrics, defineSchedules, defineSinks, defineSources, defineTask, defineThresholds, } from "./config-helpers.js";
 export type { PricingEntry, PromptEntry, SourceEntry, } from "./config-helpers.js";
 export { env } from "./env-helper.js";
+export { NoOpArtifactCollector } from "./artifact-capture/noop-collector.js";