jev-recipes 0.4.0 → 0.5.0

package/CHANGELOG.md

@@ -2,6 +2,15 @@
 
 ## Unreleased
 
+- Add `checkers-move`: supply a structured board, player, and legal moves to receive a move ID in one logical Jev request. Includes American/English checkers instructions, board translation, input checks, and offline tests, bringing the source catalog to 81 recipes. No new runtime dependencies; live speed and playing strength remain unmeasured.
+
+- Add `take-turn` and `choose-action` for narrative turn eligibility and game action recommendations, with optional ordered player history, offline tests, and a gameplay composition guide. The source catalog now contains 80 recipes.
+- Clarify that `.mjs` is a standalone Node quickstart option; existing JavaScript ES module and TypeScript projects can use their normal file types and start commands.
+
+- Add six annotation recipes: `response-refusal`, `uncertainty-expression`, `evaluation-mention`, `attribution-match`, `question-leading`, and `evidence-independence`, bringing the source catalog to 78 recipes.
+- Document their decision boundaries and research limitations; add offline contract, review-policy, discovery, and package checks without new runtime dependencies.
+- Fix the package checker to accept both legacy npm archive reports and the npm 12 object format, including dependency archives.
+
 - Add six focused recipes: `instruction-conflict`, `task-dependency`, `task-duplicate`, `constraint-strength`, `requirement-testability`, and `claim-stance`.
 - Add a catalog growth roadmap, an AI alignment research protocol, research discovery tags, and an offline first-use command in the README.
 

package/README.md

@@ -1,46 +1,45 @@
 # jev-recipes
 
-[![npm version](https://img.shields.io/npm/v/jev-recipes?logo=npm)](https://www.npmjs.com/package/jev-recipes)
+**Small AI decisions. Ready to use in your code.**
 
-<!-- BEGIN GENERATED: summary -->
+[![npm version](https://img.shields.io/npm/v/jev-recipes)](https://www.npmjs.com/package/jev-recipes)
+[![CI](https://github.com/agencyenterprise/jev-recipes/actions/workflows/ci.yml/badge.svg?branch=main)](https://github.com/agencyenterprise/jev-recipes/actions/workflows/ci.yml)
+[![Node.js version](https://img.shields.io/node/v/jev-recipes)](https://nodejs.org/)
+[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](https://github.com/agencyenterprise/jev-recipes/blob/main/LICENSE)
+
+[Quickstart](#use-a-recipe) | [Recipe catalog](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/README.md) | [API vs. SDK vs. recipes](https://github.com/agencyenterprise/jev-recipes/blob/main/docs/api-sdk-recipes.md) | [Contributing](https://github.com/agencyenterprise/jev-recipes/blob/main/CONTRIBUTING.md)
 
-72 small TypeScript recipes for decisions inside an AI application. Route a request, select useful evidence, or check a claim with a function call.
+<!-- BEGIN GENERATED: summary -->
 
-[Browse all 72 recipes](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/README.md).
+81 focused recipes for JavaScript and TypeScript. Route messages, check evidence, and label model responses with a function call.
 
 <!-- END GENERATED: summary -->
 
-[Jev](https://docs.typesafe.ai/introduction) makes the underlying decisions. This package supplies focused instructions, validated inputs and results, and clear review outcomes. Your application decides what to do next.
+Each recipe accepts your data, calls [Jev through TypeSafe's API](https://docs.typesafe.ai/introduction), and returns a structured decision. Use it in a Node.js backend, a script, or a research evaluation. Your application decides what happens next.
 
-## Try a decision
+For example, give `route` a support message and descriptions of your teams. It returns a team such as `billing`, or a review outcome when the choice is uncertain.
 
-With Node.js 22.9 or newer, run a saved illustration without an API key:
+## Watch Jev play checkers against Jev using the checkers-move recipe
 
-```sh
-npx jev-recipes demo action-scope
-```
-
-It compares a request to explain a deployment failure with a proposed production deployment and returns `additional_work`. This is an offline fixture, not a live model assessment. To evaluate your own input, install the package and set a key below.
+https://github.com/user-attachments/assets/4ac10e7b-b57c-4d73-a2fe-fee90cd1e58d
 
 ## Use a recipe
 
-Requires Node.js 22.9 or newer. Install in your application:
+Requires **Node.js 22.9 or newer**, ES modules, and a **[TypeSafe API key](https://console.typesafe.ai/keys)**. Live calls send the supplied input to TypeSafe and use API quota.
+
+**1. Install** in your project folder.
 
 ```sh
 npm install jev-recipes
 ```
 
-Set your API key in your server environment:
+**2. Add the code below to your app**, or save it as `route-message.mjs` in the same folder to try a standalone Node example.
 
-```sh
-export TYPESAFE_API_KEY='your-api-key'
-```
-
-Choose a recipe, import it, and call it:
+The `.mjs` extension is optional. We suggest it for this example because Node recognizes it as an ES module without changing your project settings. Existing apps can use `.js` with `"type": "module"` in `package.json`, or `.ts` with their usual TypeScript setup. See [Node's module formats](https://nodejs.org/api/packages.html#type).
 
 <!-- BEGIN GENERATED: quickstart -->
 
-```ts
+```js
 import { route } from 'jev-recipes/route';
 
 const result = await route({
@@ -50,26 +49,86 @@ const result = await route({
     technical: 'Errors, outages, and broken integrations',
   },
 });
-console.log(result.status, result.route);
+
+if (result.status === 'ready') {
+  console.log('Send this message to:', result.route);
+} else {
+  console.log('Needs review: ask for more detail or send to a person.');
+}
 ```
 
 <!-- END GENERATED: quickstart -->
 
-For `route`, a `ready` result contains the chosen route. A `review` result has a null route: ask for clarification or involve a person. Invalid input, malformed responses, and provider failures throw errors.
+**3. Set your key and run.** Replace `your-api-key` with your TypeSafe key. For the standalone example:
+
+```sh
+export TYPESAFE_API_KEY='your-api-key'
+node route-message.mjs
+```
+
+If you saved it as `route-message.js` in an ES module project, run `node route-message.js`. For an existing app or TypeScript project, use its normal start command.
+
+<details>
+<summary>Windows PowerShell</summary>
+
+```powershell
+$env:TYPESAFE_API_KEY = 'your-api-key'
+node route-message.mjs
+```
+
+</details>
+
+Example output:
+
+```text
+Send this message to: billing
+```
+
+A live result can differ. When the choice is uncertain, the script prints the review message instead.
 
-Live calls send the supplied input to TypeSafe and use API quota. Keep the key on the server. No Makefile or build step is needed to use the installed package.
+**Make it yours:** replace `request` with your incoming message and `routes` with your team's names and descriptions. Replace `console.log` with your queue or handler logic. Keep the key and live calls on the server.
 
-## Find the right recipe
+<a id="try-a-decision"></a>
 
-| I want to…                                     | Start with                                                                                                       |
-| ---------------------------------------------- | ---------------------------------------------------------------------------------------------------------------- |
-| Send a request to the right handler            | [`route`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/route/README.md)                     |
-| Select relevant passages                       | [`rerank`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/rerank/README.md)                   |
-| Check whether I have enough evidence to answer | [`answerability`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/answerability/README.md)     |
-| Check claims against evidence                  | [`verify`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/verify/README.md)                   |
-| Check whether a draft answers each question    | [`answer-coverage`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/answer-coverage/README.md) |
+<details>
+<summary>Try the saved example without an API key</summary>
 
-Search the catalog and inspect a recipe without an API key:
+```sh
+npx jev-recipes demo route
+```
+
+This runs an offline fixture and prints JSON with `result.status: "ready"` and `result.route: "billing"`. No model is called. Use it to inspect the interface before setting up a key.
+
+</details>
+
+## Why recipes?
+
+The TypeSafe SDK handles API calls and typed answers. Recipes add the instructions, input and response validation, confidence policy, and result handling for a specific decision.
+
+| Start with       | You provide                                                                |
+| ---------------- | -------------------------------------------------------------------------- |
+| **Raw API**      | HTTP handling, question, choices, validation, and review logic             |
+| **TypeSafe SDK** | Question, choices, task-specific validation, and review logic              |
+| **jev-recipes**  | `route({ request, routes })` and the application code that uses its result |
+
+[See the same task implemented all three ways](https://github.com/agencyenterprise/jev-recipes/blob/main/docs/api-sdk-recipes.md). All three use the same service. Recipes reduce the decision logic you need to build, test, and maintain.
+
+<a id="find-the-right-recipe"></a>
+
+## Find your recipe
+
+| Your task                                              | Start with                                                                                                   |
+| ------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------ |
+| Send a request to the right team                       | [`route`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/route/README.md)                 |
+| Find useful passages                                   | [`rerank`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/rerank/README.md)               |
+| Check whether the evidence is enough to answer         | [`answerability`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/answerability/README.md) |
+| Check claims against supplied evidence                 | [`verify`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/verify/README.md)               |
+| Find missing or ambiguous requirements                 | [`clarify`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/clarify/README.md)             |
+| Label a response's stance toward a claim               | [`claim-stance`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/claim-stance/README.md)   |
+| Choose a move from available game actions              | [`choose-action`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/choose-action/README.md) |
+| Choose a checkers move from your board and legal moves | [`checkers-move`](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/checkers-move/README.md) |
+
+Search, inspect inputs, and try saved results without a key:
 
 ```sh
 npx jev-recipes list "enough evidence" --limit 5
@@ -77,42 +136,70 @@ npx jev-recipes describe answerability
 npx jev-recipes demo answerability
 ```
 
-`list` ranks matching recipes and supports `--category`. `describe` shows when to use a recipe, related alternatives, input and result schemas, and example input. `demo` runs a saved illustration without calling a model. CLI output is JSON, so it also works in scripts and agent tools.
+[Browse the complete catalog](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/README.md). Each guide includes an import, input reference, result behavior, limitations, and related recipes.
+
+For games, the [gameplay guide](https://github.com/agencyenterprise/jev-recipes/blob/main/docs/gameplay.md) starts with a checkers decision using just the board, player, and legal moves. It also covers general action selection and narrative turn eligibility.
 
-## Run your own input from the terminal
+<a id="run-your-own-input-from-the-terminal"></a>
+
+## Use the terminal
+
+Create a JSON input file:
 
 ```sh
 npx jev-recipes example route > input.json
 ```
 
-Edit the file, then run it with your key set:
+Edit `request` and `routes` in `input.json`, then run with `TYPESAFE_API_KEY` set:
 
 ```sh
 npx jev-recipes run route input.json
 ```
 
-The installed CLI reads its environment; it does not load `.env` automatically. Errors go to stderr with exit code 1. A review outcome is a completed evaluation, so inspect the result before acting.
+Read `result.status` and `result.route` in the JSON output. Substitute another recipe's name in both commands to use a different decision. Run `npx jev-recipes --help` for all commands.
+
+The CLI reads environment variables; it does not automatically load `.env`. Errors go to stderr with exit code `1`. A review outcome is a completed evaluation, so inspect the result before acting.
+
+<a id="understand-the-result"></a>
+
+## Work with results
+
+| Result or condition                                    | What it means                                                                                                |
+| ------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------ |
+| `route`: `status: 'ready'`                             | `route` contains the chosen name. Your application can use it to choose a handler.                           |
+| `route`: `status: 'review'`                            | `route` is `null`. Ask for clarification or involve a person.                                                |
+| Other recipe results                                   | Inspect the verdict and per-item checks. A confident result can identify a conflict or an unsupported claim. |
+| Invalid input, malformed response, or provider failure | The function throws. Handle errors in your application.                                                      |
 
-## Understand the result
+Batch recipes expose their own summaries, such as `verify.allSupported` and its per-claim checks. Every result includes model and token usage. Confidence is a signal for your policy, not a guarantee of correctness.
 
-- Read the recipe's outcome as well as its confidence or review status. A confident assessment can identify an unsupported claim or missing information.
-- Batch recipes preserve individual checks. `verify` includes `allSupported` and per-claim statuses rather than one overall status.
-- Results include model and token usage. Confidence is not a guarantee of correctness.
-- Pass `{ client, model, signal }` as a second argument when you need to configure a call.
+Pass `{ client, model, signal }` as the optional second argument to configure a call. [Shared behavior and limits](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/README.md#shared-options-and-behavior) covers configuration, input limits, and uncertainty.
 
-Use direct imports such as `jev-recipes/route` to load a recipe and its dependencies. Root imports remain supported. Installation still downloads one package; direct imports do not selectively download files.
+## Verification and scope
 
-Your application owns retrieval, generation, storage, and actions. Each recipe returns a decision. See [shared behavior and limits](https://github.com/agencyenterprise/jev-recipes/blob/main/recipes/README.md#shared-options-and-behavior).
+The [CI workflow](https://github.com/agencyenterprise/jev-recipes/actions/workflows/ci.yml) runs on Node.js 22 and 24. `make ci` checks types, formatting, generated-file consistency, recipe test coverage, tooling, and the installable package.
+
+- **Package checks:** create an npm archive, install it into a separate project, and exercise imports, declarations, and offline commands.
+- **Recipe checks:** validate software behavior using mocked responses, including invalid inputs, confidence boundaries, and malformed answers.
+- **Model evaluation:** offline tests and demos do not measure Jev's accuracy. Evaluate recipes on your own data for your intended use.
+
+The package contains compiled code, type declarations, catalog data, and offline demo assets. Tests, build tools, and source guides stay out of the npm archive. [Publishing details](https://github.com/agencyenterprise/jev-recipes/blob/main/RELEASING.md).
+
+Direct imports such as `jev-recipes/route` load the selected recipe and its dependencies. Installation downloads one package; it does not selectively download individual recipes.
 
 ## AI alignment research
 
-Use focused decisions to annotate model outputs for research. `claim-stance` labels a response's expressed agreement with a claim; `verify` checks supplied evidence; `draft-compare` compares responses under a rubric. These are candidate components of an evaluation, with uncertainty available for review.
+Recipes can serve as candidate annotation tools in controlled experiments. `claim-stance` labels expressed agreement, `verify` checks supplied evidence, and `draft-compare` compares responses under a rubric.
 
-The [AI alignment research guide](https://github.com/agencyenterprise/jev-recipes/blob/main/docs/ai-alignment-research.md) outlines a controlled agreement study, human annotation checks, and ways to work with saved outputs. The recipes do not infer a model's internal motives or establish that it is aligned.
+The [research guide](https://github.com/agencyenterprise/jev-recipes/blob/main/docs/ai-alignment-research.md) outlines a controlled agreement study and validation against human annotations. These labels describe observable outputs; they do not establish internal motives or prove that a model is aligned.
 
-## Contribute
+<a id="contribute"></a>
 
-The recipe folder owns its code, schemas, metadata, demo, and guide. Tests stay under `tests/recipe/` and are not published. Exports and the catalog are generated from the recipe folders.
+## Contributing and support
+
+[Report a bug or request a recipe](https://github.com/agencyenterprise/jev-recipes/issues). Include the recipe name, package and Node.js versions, and a minimal reproduction. Keep API keys and private inputs out of reports.
+
+To work on the repository:
 
 ```sh
 make setup
@@ -120,7 +207,9 @@ make docs
 make ci
 ```
 
-Run `make help` for all commands. [Contributing](https://github.com/agencyenterprise/jev-recipes/blob/main/CONTRIBUTING.md) explains authoring and generation; [releasing](https://github.com/agencyenterprise/jev-recipes/blob/main/RELEASING.md) explains package checks and publishing.
+The recipe folder owns its implementation, schemas, metadata, demo, and guide. Exports and catalog entries are generated from those folders. Tests live under `tests/recipe/`.
+
+[Contributing guide](https://github.com/agencyenterprise/jev-recipes/blob/main/CONTRIBUTING.md) | [Roadmap](https://github.com/agencyenterprise/jev-recipes/blob/main/RECIPE_ROADMAP.md) | [Changelog](https://github.com/agencyenterprise/jev-recipes/blob/main/CHANGELOG.md)
 
 ## License
 

package/dist/catalog/generated/details/attribution-match.json

@@ -0,0 +1,55 @@
+{
+  "inputSchema": {
+    "$schema": "https://json-schema.org/draft/2020-12/schema",
+    "type": "object",
+    "properties": {
+      "statement": {
+        "type": "string",
+        "description": "One statement or faithful paraphrase whose attributed speaker or source should be checked."
+      },
+      "attributedTo": {
+        "type": "string",
+        "description": "The speaker or source claimed to have made or explicitly endorsed the statement."
+      },
+      "source": {
+        "type": "string",
+        "description": "The supplied transcript or excerpt, including speaker labels or attribution context needed to check the claim."
+      },
+      "minConfidence": { "type": "number", "minimum": 0, "maximum": 1 }
+    },
+    "required": ["statement", "attributedTo", "source"]
+  },
+  "resultSchema": {
+    "$schema": "https://json-schema.org/draft/2020-12/schema",
+    "type": "object",
+    "properties": {
+      "model": { "type": "string" },
+      "usage": {
+        "type": "object",
+        "properties": {
+          "input_tokens": { "type": "integer", "minimum": 0, "maximum": 9007199254740991 },
+          "output_tokens": { "type": "integer", "minimum": 0, "maximum": 9007199254740991 }
+        },
+        "required": ["input_tokens", "output_tokens"],
+        "additionalProperties": false
+      },
+      "status": { "type": "string", "enum": ["ready", "review"] },
+      "verdict": {
+        "type": "string",
+        "enum": ["matched", "mismatched", "not_attributed", "unclear"]
+      },
+      "confidence": { "type": "number", "minimum": 0, "maximum": 1 },
+      "probabilities": {
+        "type": "object",
+        "propertyNames": {
+          "type": "string",
+          "enum": ["matched", "mismatched", "not_attributed", "unclear"]
+        },
+        "additionalProperties": { "type": "number", "minimum": 0, "maximum": 1 },
+        "required": ["matched", "mismatched", "not_attributed", "unclear"]
+      }
+    },
+    "required": ["model", "usage", "status", "verdict", "confidence", "probabilities"],
+    "additionalProperties": false
+  }
+}

package/dist/catalog/generated/details/checkers-move.json

@@ -0,0 +1,128 @@
+{
+  "inputSchema": {
+    "$schema": "https://json-schema.org/draft/2020-12/schema",
+    "type": "object",
+    "properties": {
+      "board": {
+        "minItems": 1,
+        "maxItems": 24,
+        "type": "array",
+        "items": {
+          "type": "object",
+          "properties": {
+            "square": {
+              "type": "string",
+              "pattern": "^(?:[aceg][1357]|[bdfh][2468])$",
+              "description": "A playable square: a1 is bottom left; files a-h, ranks 1-8."
+            },
+            "player": { "type": "string", "enum": ["red", "black"] },
+            "king": {
+              "description": "True for a king; omitted or false for a man.",
+              "type": "boolean"
+            }
+          },
+          "required": ["square", "player"],
+          "additionalProperties": false
+        },
+        "description": "All remaining pieces. Unlisted squares are empty; red moves toward rank 8, black toward rank 1."
+      },
+      "player": {
+        "type": "string",
+        "enum": ["red", "black"],
+        "description": "The side whose turn it is and whose move should be chosen."
+      },
+      "legalMoves": {
+        "minItems": 1,
+        "maxItems": 50,
+        "type": "array",
+        "items": {
+          "type": "object",
+          "properties": {
+            "id": {
+              "type": "string",
+              "description": "Your move ID, returned unchanged when selected. __proto__ is reserved."
+            },
+            "from": {
+              "type": "string",
+              "pattern": "^(?:[aceg][1357]|[bdfh][2468])$",
+              "description": "A playable square: a1 is bottom left; files a-h, ranks 1-8."
+            },
+            "path": {
+              "minItems": 1,
+              "maxItems": 12,
+              "type": "array",
+              "items": {
+                "type": "string",
+                "pattern": "^(?:[aceg][1357]|[bdfh][2468])$",
+                "description": "A playable square: a1 is bottom left; files a-h, ranks 1-8."
+              },
+              "description": "Every landing square in order, excluding from. Include the full jump sequence."
+            },
+            "captures": {
+              "description": "Captured piece squares in jump order. Omit or use [] for an ordinary move.",
+              "maxItems": 12,
+              "type": "array",
+              "items": {
+                "type": "string",
+                "pattern": "^(?:[aceg][1357]|[bdfh][2468])$",
+                "description": "A playable square: a1 is bottom left; files a-h, ranks 1-8."
+              }
+            }
+          },
+          "required": ["id", "from", "path"],
+          "additionalProperties": false
+        },
+        "description": "One to fifty complete legal moves from your game, with unique IDs."
+      },
+      "minConfidence": { "type": "number", "minimum": 0, "maximum": 1 }
+    },
+    "required": ["board", "player", "legalMoves"],
+    "additionalProperties": false
+  },
+  "resultSchema": {
+    "$schema": "https://json-schema.org/draft/2020-12/schema",
+    "type": "object",
+    "properties": {
+      "model": { "type": "string" },
+      "usage": {
+        "type": "object",
+        "properties": {
+          "input_tokens": { "type": "integer", "minimum": 0, "maximum": 9007199254740991 },
+          "output_tokens": { "type": "integer", "minimum": 0, "maximum": 9007199254740991 }
+        },
+        "required": ["input_tokens", "output_tokens"],
+        "additionalProperties": false
+      },
+      "status": { "type": "string", "enum": ["ready", "review"] },
+      "verdict": { "type": "string", "enum": ["matched", "none", "ambiguous"] },
+      "selection": { "type": ["string", "null"] },
+      "suggestedSelection": { "type": ["string", "null"] },
+      "confidence": { "type": "number", "minimum": 0, "maximum": 1 },
+      "probabilities": {
+        "type": "object",
+        "properties": {
+          "candidates": {
+            "type": "object",
+            "propertyNames": { "type": "string" },
+            "additionalProperties": { "type": "number", "minimum": 0, "maximum": 1 }
+          },
+          "none": { "type": "number", "minimum": 0, "maximum": 1 },
+          "ambiguous": { "type": "number", "minimum": 0, "maximum": 1 }
+        },
+        "required": ["candidates", "none", "ambiguous"],
+        "additionalProperties": false
+      }
+    },
+    "required": [
+      "model",
+      "usage",
+      "status",
+      "verdict",
+      "selection",
+      "suggestedSelection",
+      "confidence",
+      "probabilities"
+    ],
+    "additionalProperties": false
+  }
+}

package/dist/catalog/generated/details/choose-action.json

@@ -0,0 +1,102 @@
+{
+  "inputSchema": {
+    "$schema": "https://json-schema.org/draft/2020-12/schema",
+    "type": "object",
+    "properties": {
+      "player": {
+        "type": "string",
+        "description": "The player whose next action should be chosen."
+      },
+      "objective": {
+        "type": "string",
+        "description": "The goal or scoring preference used to compare eligible actions."
+      },
+      "rules": {
+        "type": "string",
+        "description": "Applicable game rules, including turn, phase, resource, and action constraints."
+      },
+      "environment": {
+        "type": "string",
+        "description": "The current game snapshot available to the player, after the supplied history."
+      },
+      "actions": {
+        "minItems": 1,
+        "maxItems": 50,
+        "type": "array",
+        "items": {
+          "type": "object",
+          "properties": { "id": { "type": "string" }, "text": { "type": "string" } },
+          "required": ["id", "text"]
+        },
+        "description": "One to fifty candidate actions with unique IDs and concrete action descriptions."
+      },
+      "history": {
+        "maxItems": 100,
+        "type": "array",
+        "items": {
+          "type": "object",
+          "properties": {
+            "player": {
+              "type": "string",
+              "description": "The player who took the recorded action."
+            },
+            "action": {
+              "type": "string",
+              "description": "The observed action, including its known outcome when relevant."
+            }
+          },
+          "required": ["player", "action"]
+        },
+        "description": "Optional observed actions by any players, oldest first. May be empty or incomplete."
+      },
+      "minConfidence": { "type": "number", "minimum": 0, "maximum": 1 }
+    },
+    "required": ["player", "objective", "rules", "environment", "actions"]
+  },
+  "resultSchema": {
+    "$schema": "https://json-schema.org/draft/2020-12/schema",
+    "type": "object",
+    "properties": {
+      "model": { "type": "string" },
+      "usage": {
+        "type": "object",
+        "properties": {
+          "input_tokens": { "type": "integer", "minimum": 0, "maximum": 9007199254740991 },
+          "output_tokens": { "type": "integer", "minimum": 0, "maximum": 9007199254740991 }
+        },
+        "required": ["input_tokens", "output_tokens"],
+        "additionalProperties": false
+      },
+      "status": { "type": "string", "enum": ["ready", "review"] },
+      "verdict": { "type": "string", "enum": ["matched", "none", "ambiguous"] },
+      "selection": { "type": ["string", "null"] },
+      "suggestedSelection": { "type": ["string", "null"] },
+      "confidence": { "type": "number", "minimum": 0, "maximum": 1 },
+      "probabilities": {
+        "type": "object",
+        "properties": {
+          "candidates": {
+            "type": "object",
+            "propertyNames": { "type": "string" },
+            "additionalProperties": { "type": "number", "minimum": 0, "maximum": 1 }
+          },
+          "none": { "type": "number", "minimum": 0, "maximum": 1 },
+          "ambiguous": { "type": "number", "minimum": 0, "maximum": 1 }
+        },
+        "required": ["candidates", "none", "ambiguous"],
+        "additionalProperties": false
+      }
+    },
+    "required": [
+      "model",
+      "usage",
+      "status",
+      "verdict",
+      "selection",
+      "suggestedSelection",
+      "confidence",
+      "probabilities"
+    ],
+    "additionalProperties": false
+  }
+}

package/dist/catalog/generated/details/evaluation-mention.json

@@ -0,0 +1,48 @@
+{
+  "inputSchema": {
+    "$schema": "https://json-schema.org/draft/2020-12/schema",
+    "type": "object",
+    "properties": {
+      "response": {
+        "type": "string",
+        "description": "The response to inspect for explicit references to evaluation of an AI response or model."
+      },
+      "context": {
+        "type": "string",
+        "description": "Surrounding text used only to resolve who or what the response refers to."
+      },
+      "minConfidence": { "type": "number", "minimum": 0, "maximum": 1 }
+    },
+    "required": ["response"]
+  },
+  "resultSchema": {
+    "$schema": "https://json-schema.org/draft/2020-12/schema",
+    "type": "object",
+    "properties": {
+      "model": { "type": "string" },
+      "usage": {
+        "type": "object",
+        "properties": {
+          "input_tokens": { "type": "integer", "minimum": 0, "maximum": 9007199254740991 },
+          "output_tokens": { "type": "integer", "minimum": 0, "maximum": 9007199254740991 }
+        },
+        "required": ["input_tokens", "output_tokens"],
+        "additionalProperties": false
+      },
+      "status": { "type": "string", "enum": ["ready", "review"] },
+      "verdict": { "type": "string", "enum": ["self_reference", "discussion", "none", "unclear"] },
+      "confidence": { "type": "number", "minimum": 0, "maximum": 1 },
+      "probabilities": {
+        "type": "object",
+        "propertyNames": {
+          "type": "string",
+          "enum": ["self_reference", "discussion", "none", "unclear"]
+        },
+        "additionalProperties": { "type": "number", "minimum": 0, "maximum": 1 },
+        "required": ["self_reference", "discussion", "none", "unclear"]
+      }
+    },
+    "required": ["model", "usage", "status", "verdict", "confidence", "probabilities"],
+    "additionalProperties": false
+  }
+}