@zhuoyuezs/ml-platform 0.2.3-alpha → 0.2.3-alpha.1

package/DEVELOPMENT.md

@@ -10,7 +10,7 @@
 - Skill 直接调用正式命令 `ml-platform`，不嵌入 CLI runtime。
 - 正式 npm 包只有 `@zhuoyuezs/ml-platform`，正式 bin 只有 `ml-platform`。
 - `business-client-js/src-ts/` 是 business CLI 的唯一 TypeScript 源码，`src/` 是供 Node.js 和
-  release payload 使用的编译输出；仓库 `skills/` 下四个受支持 Skill 是各自唯一源码。
+  release payload 使用的编译输出；仓库 `skills/` 下五个受支持 Skill 是各自唯一源码。
 - `npx @zhuoyuezs/ml-platform@<version> install` 是面向用户的统一安装入口；默认目标是 Codex
   用户级 Skill，不要求额外安装参数。
 - 管理命令默认输出简洁结果；Agent 和 CI 显式使用 `--json` 获取稳定结构化输出。
@@ -34,6 +34,7 @@ skills/feature-management/
 skills/data-source-exploration/
 skills/model-lifecycle-management/
 skills/model-deployment-management/
+skills/prediction-report-management/
 npm/ml-platform/
 npm/scripts/build_ml_platform_release.py
 agent_poc/codex-host/

package/README.md

@@ -1,7 +1,8 @@
 # ML Platform
 
 ML Platform 的 TypeScript CLI（编译为 Node.js JavaScript runtime）、`feature-management`、
-`data-source-exploration`、`model-lifecycle-management` 与 `model-deployment-management`
+`data-source-exploration`、`model-lifecycle-management`、`model-deployment-management` 与
+`prediction-report-management`
 Agent Skill。CLI 与 Skill 由同一个 npm
 release 安装和升级，运行时不需要 Python 或 `uv`。
 
@@ -19,7 +20,7 @@ release 安装和升级，运行时不需要 Python 或 `uv`。
 npx @zhuoyuezs/ml-platform@latest install
 ```
 
-这条命令不需要额外参数。默认安装持久化的 `ml-platform` 命令，并将四个 Skill 安装到
+这条命令不需要额外参数。默认安装持久化的 `ml-platform` 命令，并将五个 Skill 安装到
 `~/.codex/skills/`，供当前用户的 Codex 使用。安装后重启终端和 Agent 会话。
 
 只在当前项目使用 Skill 时，在项目根目录运行：
@@ -131,8 +132,9 @@ npx --yes --registry=https://registry.npmjs.org/ \
 ## 包内容
 
 - 一份由 TypeScript 编译的 business CLI；
-- 四份纯 Skill：`feature-management`、`data-source-exploration`、
-  `model-lifecycle-management` 与 `model-deployment-management`；
+- 五份纯 Skill：`feature-management`、`data-source-exploration`、
+  `model-lifecycle-management`、`model-deployment-management` 与
+  `prediction-report-management`；
 - 显式安装、升级、状态、校验和回滚逻辑。
 
 不包含 Codex CLI、Python business-client wheel、Agent Host、平台服务端、模型训练或 Kubernetes

package/checksums.json

@@ -2,8 +2,8 @@
   "files": [
     {
       "path": "runtime/business-client/README.md",
-      "sha256": "sha256:c263ebe1591d38d90fb68227f42c6de43014a05e98b736d2e3363e19545676d9",
-      "size_bytes": 8779
+      "sha256": "sha256:f03d10f0de78e4ff456a6718b5b74b7fd001598f5a5963985d97d0b99403c0a5",
+      "size_bytes": 12756
     },
     {
       "path": "runtime/business-client/package-lock.json",
@@ -22,8 +22,8 @@
     },
     {
       "path": "runtime/business-client/src/cli.js",
-      "sha256": "sha256:739be8665f611802d9196ba84566578fbfc6544c6907af165edbaf115d9d1998",
-      "size_bytes": 62651
+      "sha256": "sha256:ecdc7d2e93323a4dd04514573219061f54db6583d955dfa82afc482abcd3335a",
+      "size_bytes": 66532
     },
     {
       "path": "runtime/business-client/src/config.js",
@@ -32,13 +32,13 @@
     },
     {
       "path": "runtime/business-client/src/http.js",
-      "sha256": "sha256:3c8b40b0da0bb87c95f1367f540cd08a7ec28e0d5fe8e52a0367fecb5ae4094c",
-      "size_bytes": 18763
+      "sha256": "sha256:517ad46944f3a4068a00a570dc09975c5b59de61b730e0acf8e009ee1f18c89d",
+      "size_bytes": 18865
     },
     {
       "path": "skills/data-source-exploration/SKILL.md",
-      "sha256": "sha256:3b609d61f3c73da3657ce2e6bb4d1eff95a90e4ac5e5445b8513536438b26b0b",
-      "size_bytes": 3853
+      "sha256": "sha256:11b20671ad81b42cd2ddb1cb146196b9f5ab15e962249e7e07cfa148c601e881",
+      "size_bytes": 4172
     },
     {
       "path": "skills/data-source-exploration/agents/openai.yaml",
@@ -47,8 +47,8 @@
     },
     {
       "path": "skills/feature-management/SKILL.md",
-      "sha256": "sha256:309d2ae292113f99bd17b6865b87bbe793809c88005458f477fa1e80c41b1523",
-      "size_bytes": 33334
+      "sha256": "sha256:1118aef3734fe14fd8f7d1074ff730340663e42565155bde984ab5b397c2da77",
+      "size_bytes": 33777
     },
     {
       "path": "skills/feature-management/agents/openai.yaml",
@@ -127,8 +127,8 @@
     },
     {
       "path": "skills/model-deployment-management/SKILL.md",
-      "sha256": "sha256:d50f64fa709c8ff36da97c60ce0d20b1217d1d9f78f602048f49c0c4bbab7ea2",
-      "size_bytes": 3927
+      "sha256": "sha256:e8419e4d83031248b40783674813bf3f5a98544f05ea3fbe056547ae47632016",
+      "size_bytes": 4397
     },
     {
       "path": "skills/model-deployment-management/agents/openai.yaml",
@@ -152,8 +152,8 @@
     },
     {
       "path": "skills/model-lifecycle-management/SKILL.md",
-      "sha256": "sha256:1f2174fac7f82e7539c781b1c7bc4164a8adc393acb33b3b2165408b25600135",
-      "size_bytes": 4888
+      "sha256": "sha256:9ad2428621b52a15a83dce34a2a95606d7d979017a227c0aeb8fe5c8acd5faa3",
+      "size_bytes": 5442
     },
     {
       "path": "skills/model-lifecycle-management/agents/openai.yaml",
@@ -189,6 +189,16 @@
       "path": "skills/model-lifecycle-management/references/zero-shot-inference.md",
       "sha256": "sha256:4ea224be54a7b0fc7754bdaa41df0f3b14c56f5134afc3951e91e59844108f90",
       "size_bytes": 2350
+    },
+    {
+      "path": "skills/prediction-report-management/SKILL.md",
+      "sha256": "sha256:32abad9d9a152680d126cf1fa3df921a367a9ad9eb0aaaf3ce48b05538b4fbd5",
+      "size_bytes": 4365
+    },
+    {
+      "path": "skills/prediction-report-management/agents/openai.yaml",
+      "sha256": "sha256:fd86629700aa9a4f1ab985d87883162ab260f469d5059a516abcd1e3273d89fa",
+      "size_bytes": 241
     }
   ],
   "schema_version": "data_platform.ml_platform_checksums/v1"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@zhuoyuezs/ml-platform",
-  "version": "0.2.3-alpha",
+  "version": "0.2.3-alpha.1",
   "description": "ML 数据平台 business CLI 与 Agent Skills 管理工具",
   "bin": {
     "ml-platform": "scripts/main.js"

package/release.json

@@ -13,7 +13,7 @@
     "entrypoint": "src/cli.js",
     "name": "ml-platform",
     "path": "runtime/business-client",
-    "sha256": "sha256:ff1b5f6de8111ac9dea00701afadbb6e47cc14937fa78b958ec8a3b777a4ecb9",
+    "sha256": "sha256:eaa88929b414e386294287009b4ecb25e26d1eb39403469ef649af145b2208de",
     "version": "0.8.3"
   },
   "policy_sha256": "sha256:02fbde0134696b78c0c365d6e0b76c9814227208da5b2b3d4e5dbd82dc4a3606",
@@ -25,7 +25,7 @@
     "ml_data_platform.feature_set/v1",
     "ml_data_platform.dataset_manifest/v1"
   ],
-  "release_version": "0.2.3-alpha",
+  "release_version": "0.2.3-alpha.1",
   "runtime_requirements": {
     "node": ">=18",
     "os": [
@@ -38,27 +38,33 @@
     "data-source-exploration": {
       "path": "skills/data-source-exploration",
       "requires_cli": ">=0.8.3 <0.9.0",
-      "revision": "0.2.3-alpha",
-      "sha256": "sha256:ae3cfc88f8f3f46a5b7e0179d850b2216bc37275f34679dbea1ff2ba99875a96"
+      "revision": "0.2.3-alpha.1",
+      "sha256": "sha256:2fa1e1802d799ee0c47b10263f2fde57b604e505fea5c98bfa23f51f382bbf69"
     },
     "feature-management": {
       "path": "skills/feature-management",
       "requires_cli": ">=0.8.3 <0.9.0",
-      "revision": "0.2.3-alpha",
-      "sha256": "sha256:63e3bc337bc4cc4cdd38bfdd99ff9d8fd6c7d57d24aec1b7f70101ce38b1f599"
+      "revision": "0.2.3-alpha.1",
+      "sha256": "sha256:6dfa708172c3b5b23af75997c82a6d43bc2eb8ca67cc2275c8b95022455b0c42"
     },
     "model-deployment-management": {
       "path": "skills/model-deployment-management",
       "requires_cli": ">=0.8.3 <0.9.0",
-      "revision": "0.2.3-alpha",
-      "sha256": "sha256:e7d87c272a0dcc028ea628d5e36f5b7e52ed27cb9c1bc17ea76527478e28b46d"
+      "revision": "0.2.3-alpha.1",
+      "sha256": "sha256:d3f3fe0e4a11b93d8f6f57758164aa236b84a49c1b9f0d0b700fd9d6d615a8d7"
     },
     "model-lifecycle-management": {
       "path": "skills/model-lifecycle-management",
       "requires_cli": ">=0.8.3 <0.9.0",
-      "revision": "0.2.3-alpha",
-      "sha256": "sha256:13f94cb0928d15a83b8bb10195dc3a83c28c4bb7723bc8a35e262b6c1ba46b0d"
+      "revision": "0.2.3-alpha.1",
+      "sha256": "sha256:8822fbec7037c2312d18b1d18648b60757fc569e6bf5f2856ea41592f3ba987d"
+    },
+    "prediction-report-management": {
+      "path": "skills/prediction-report-management",
+      "requires_cli": ">=0.8.3 <0.9.0",
+      "revision": "0.2.3-alpha.1",
+      "sha256": "sha256:a0210cd93244bce41e5b71f8dc70c6c4a79973968cd22af7e0e0102eae3efaa3"
     }
   },
-  "source_commit": "42714d93a4fdcc29fce8d98bc8e17823535cb56f"
+  "source_commit": "07c13464f3ed2c10644ba6707d329b54f48e725d"
 }

package/runtime/business-client/README.md

@@ -73,6 +73,91 @@ ml-platform get-deployment-events <deployment-id> --limit 100 --offset 0
 `list-deployments` 也支持按 `service`、`environment`、`state` 和自由文本 `q` 过滤，返回平台
 Deployment Controller 管理的部署，不把模型中心回写的 delivery 状态当作平台部署。
 
+## Prediction reporting
+
+定时预测上报通过业务 CLI 管理版本化推送目标、计划与运行记录。下面的 JSON 是字段示例，
+不是现网配置；替换目标地址、响应字段和占位凭据后，把实际配置文件放在仓库外并限制读取权限。
+不要把认证信息写入 Kubernetes Secret、命令行参数或 Git。`validate-destination-profile`
+仅做服务端合同校验，不会访问 Token 或推送接口。
+
+`PROFILE_JSON` 示例：
+
+```json
+{
+  "schema_version": "ml_data_platform.destination_profile/v1",
+  "name": "forecast_sink",
+  "version": "v1",
+  "environment": "test",
+  "base_url": "https://integration.example.invalid/api",
+  "path": "/forecast/report",
+  "payload_mapping": "forecast-report-v1",
+  "authentication": {
+    "kind": "token_endpoint",
+    "token_request": {
+      "method": "GET",
+      "path": "/auth/getToken",
+      "parameters": [
+        {"name": "appKey", "value": "<app-key>", "location": "query"},
+        {"name": "appSecurity", "value": "<app-secret>", "location": "query"}
+      ]
+    },
+    "success": {"http_status": [200], "body_code_path": "code", "body_code_values": [200]},
+    "token_path": "data.token",
+    "header_name": "Authorization",
+    "header_scheme": "Bearer"
+  },
+  "response_success": {"http_status": [200], "body_code_path": "code", "body_code_values": [200]},
+  "max_token_refreshes_per_day": 20
+}
+```
+
+`token_path`、业务成功码和 `/forecast/report` 只是示例，必须按目标接口的真实响应和路径配置。
+若响应提供有效期，可设置 `expires_in_path` 与 `expires_in_unit`；否则服务端默认按 7200 秒缓存，
+并在 `refresh_skew_seconds` 前刷新。不同环境应使用不同 Profile 版本或名称；修改目标地址、
+认证规则或凭据时创建新版本。`get`/`list` 只返回凭据已配置的脱敏标记，无法取回明文。
+
+```bash
+ml-platform --profile server validate-destination-profile /absolute/path/to/profile.json
+ml-platform --profile server create-destination-profile /absolute/path/to/profile.json
+ml-platform --profile server list-destination-profiles --limit 50 --offset 0
+ml-platform --profile server get-destination-profile forecast_sink v1
+```
+
+计划引用已有 Profile，不复制凭据。`SCHEDULE_JSON` 示例：
+
+```json
+{
+  "schema_version": "ml_data_platform.prediction_report_schedule/v1",
+  "schedule_id": "forecast_hourly",
+  "name": "Hourly forecast report",
+  "deployment_id": "<deployment-id>",
+  "enabled": false,
+  "cron": "0 * * * *",
+  "timezone": "Asia/Shanghai",
+  "request": {"entity": {"site": "site-1"}, "horizon": "1h"},
+  "destination": {"name": "forecast_sink", "version": "v1"}
+}
+```
+
+计划要求部署的模型支持 `feature_lookup`；部署必须有可用推理策略，或在计划请求中提供
+`horizon`。五段 cron 至少间隔五分钟。`enabled: false` 用于先创建但不发送的计划；
+确认目标接口可用且部署 READY 后再恢复。计划创建或 Profile 校验成功均不等于已推送成功。
+
+```bash
+ml-platform --profile server create-prediction-report-schedule /absolute/path/to/schedule.json
+ml-platform --profile server list-prediction-report-schedules --limit 50 --offset 0
+ml-platform --profile server get-prediction-report-schedule forecast_hourly
+ml-platform --profile server update-prediction-report-schedule forecast_hourly /absolute/path/to/changes.json
+ml-platform --profile server pause-prediction-report-schedule forecast_hourly
+ml-platform --profile server resume-prediction-report-schedule forecast_hourly
+ml-platform --profile server list-prediction-report-runs forecast_hourly --limit 50 --offset 0
+ml-platform --profile server get-prediction-report-run <run-id>
+```
+
+`CHANGES_JSON` 是仅含变更字段的对象，例如
+`{"destination":{"name":"forecast_sink","version":"v2"}}`。更新不会改写历史运行；
+是否已投递成功以 Run 的 `status`、`destination_status_code` 和错误码为准。
+
 `READY` 必须同时包含 `ReplicasReady=true`、`ArtifactsVerified=true` 和
 `SmokePredictionPassed=true`。`get-deployment` 返回控制器实际观察到的 endpoint 元数据；
 本地 CLI 不推测 Kubernetes Service 名称。部署后验证优先使用控制面代理：

package/runtime/business-client/src/cli.js

@@ -40,6 +40,10 @@ const BUSINESS_COMMANDS = new Set([
     "get-deployment", "wait-deployment", "reconcile-deployment", "approve-deployment",
     "promote-deployment", "rollback-deployment", "suspend-deployment", "resume-deployment",
     "undeploy-deployment", "get-deployment-events", "predict-deployment",
+    "validate-destination-profile", "create-destination-profile", "list-destination-profiles", "get-destination-profile",
+    "create-prediction-report-schedule", "list-prediction-report-schedules", "get-prediction-report-schedule",
+    "update-prediction-report-schedule", "pause-prediction-report-schedule", "resume-prediction-report-schedule",
+    "list-prediction-report-runs", "get-prediction-report-run",
 ]);
 const COMMAND_USAGE = {
     ...Object.fromEntries([...LIFECYCLE_COLLECTIONS].map((name) => [`list-${name}`, `list-${name} [--limit N] [--offset N] [-q QUERY]`])),
@@ -166,6 +170,18 @@ const COMMAND_USAGE = {
     "undeploy-deployment": "undeploy-deployment DEPLOYMENT_ID --actor NAME --reason TEXT",
     "get-deployment-events": "get-deployment-events DEPLOYMENT_ID [--limit N] [--offset N]",
     "predict-deployment": "predict-deployment DEPLOYMENT_ID REQUEST_JSON",
+    "create-destination-profile": "create-destination-profile PROFILE_JSON",
+    "validate-destination-profile": "validate-destination-profile PROFILE_JSON",
+    "list-destination-profiles": "list-destination-profiles [--limit N] [--offset N]",
+    "get-destination-profile": "get-destination-profile NAME VERSION",
+    "create-prediction-report-schedule": "create-prediction-report-schedule SCHEDULE_JSON",
+    "list-prediction-report-schedules": "list-prediction-report-schedules [--limit N] [--offset N]",
+    "get-prediction-report-schedule": "get-prediction-report-schedule SCHEDULE_ID",
+    "update-prediction-report-schedule": "update-prediction-report-schedule SCHEDULE_ID CHANGES_JSON",
+    "pause-prediction-report-schedule": "pause-prediction-report-schedule SCHEDULE_ID",
+    "resume-prediction-report-schedule": "resume-prediction-report-schedule SCHEDULE_ID",
+    "list-prediction-report-runs": "list-prediction-report-runs SCHEDULE_ID [--limit N] [--offset N]",
+    "get-prediction-report-run": "get-prediction-report-run RUN_ID",
     apply: "apply CATALOG_DIRECTORY [--dry-run] [--project PROJECT]",
 };
 function usage(command) {
@@ -858,6 +874,40 @@ async function runBusinessCli(argv) {
         const request = readSpec(positional(rest, "request JSON"));
         result = await client(options).post(`/deployments/${encodeURIComponent(deploymentId)}/predict`, request);
     }
+    else if (["create-destination-profile", "validate-destination-profile"].includes(options.command)) {
+        result = await client(options).post(options.command === "create-destination-profile" ? "/destination-profiles" : "/destination-profiles/validate", readSpec(positional(rest, "profile JSON")));
+    }
+    else if (options.command === "list-destination-profiles") {
+        result = await listCollection(client(options), "/destination-profiles", {}, rest);
+    }
+    else if (options.command === "get-destination-profile") {
+        const name = positional(rest, "profile name");
+        const version = positional(rest, "profile version");
+        result = await client(options).get(`/destination-profiles/${encodeURIComponent(name)}/versions/${encodeURIComponent(version)}`);
+    }
+    else if (options.command === "create-prediction-report-schedule") {
+        result = await client(options).post("/prediction-report-schedules", readSpec(positional(rest, "schedule JSON")));
+    }
+    else if (options.command === "list-prediction-report-schedules") {
+        result = await listCollection(client(options), "/prediction-report-schedules", {}, rest);
+    }
+    else if (["get-prediction-report-schedule", "update-prediction-report-schedule", "pause-prediction-report-schedule", "resume-prediction-report-schedule", "list-prediction-report-runs"].includes(options.command)) {
+        const scheduleId = positional(rest, "schedule id");
+        const endpoint = `/prediction-report-schedules/${encodeURIComponent(scheduleId)}`;
+        if (options.command === "update-prediction-report-schedule")
+            result = await client(options).patch(endpoint, readSpec(positional(rest, "changes JSON")));
+        else if (options.command === "pause-prediction-report-schedule")
+            result = await client(options).post(`${endpoint}/pause`);
+        else if (options.command === "resume-prediction-report-schedule")
+            result = await client(options).post(`${endpoint}/resume`);
+        else if (options.command === "list-prediction-report-runs")
+            result = await listCollection(client(options), `${endpoint}/runs`, {}, rest);
+        else
+            result = await client(options).get(endpoint);
+    }
+    else if (options.command === "get-prediction-report-run") {
+        result = await client(options).get(`/prediction-report-runs/${encodeURIComponent(positional(rest, "run id"))}`);
+    }
     else if (options.command === "get-model-runtime-release") {
         result = await client(options).get("/model-runtime-release");
     }

package/runtime/business-client/src/http.js

@@ -62,6 +62,7 @@ class PlatformApiClient {
     post(endpoint, payload, params) { return this.request("POST", endpoint, { payload, params }); }
     postWithHeaders(endpoint, payload, requestHeaders) { return this.request("POST", endpoint, { payload, requestHeaders }); }
     put(endpoint, payload, params) { return this.request("PUT", endpoint, { payload, params }); }
+    patch(endpoint, payload, params) { return this.request("PATCH", endpoint, { payload, params }); }
     delete(endpoint, params) { return this.request("DELETE", endpoint, { params }); }
     uploadFoundationCsv(endpoint, schema, forecast, filePath, requestHeaders = {}) {
         const resolved = path.resolve(expandHome(filePath));

package/skills/data-source-exploration/SKILL.md

@@ -88,3 +88,8 @@ After a human confirms source mapping, event time, entity keys, units, availabil
 horizon, and feature formulas, use `$feature-management` to create governed assets and build a
 DatasetArtifact. Use `$model-lifecycle-management` only after that artifact exists. Exploration does not
 authorize publishing assets, building datasets, or submitting training jobs.
+
+At the end of an exploration task, offer the next relevant option: confirm the
+unresolved business semantics with an owner, or, when they are already confirmed,
+define governed assets with `$feature-management`. Name the evidence that makes
+the option available; do not present exploratory rows as a finished dataset.

package/skills/feature-management/SKILL.md

@@ -503,3 +503,10 @@ actual compute backend
 ```
 
 If the work stops before publication, distinguish generated files, locally validated files, dry-run validation, and remotely published assets.
+
+Close with at most two optional next steps tied to the actual checkpoint. For
+example, after a dry-run, offer publication only with separate approval; after
+a verified DatasetArtifact, offer training or evaluation through
+`$model-lifecycle-management`. If the artifact is missing or validation failed,
+name that blocker instead of suggesting model training. A suggested step does
+not authorize publishing, building, or submitting a workload.

package/skills/model-deployment-management/SKILL.md

@@ -63,3 +63,10 @@ deployment does not reach `READY` or an operation is rejected.
 Report the deployment ID, revision, immutable image digest, desired and
 observed state, failed conditions, endpoint reachability, and event evidence.
 Do not claim a deployment is usable merely because its Pod is running.
+
+Close with an optional next step based on observed state: resolve failed READY
+conditions before prediction; once READY, offer `predict-deployment` to verify
+the model. After a successful deployment-bound prediction, offer scheduled
+delivery through `$prediction-report-management` only when recurring reporting
+is relevant and the package supports `feature_lookup`. Do not create a report
+schedule or imply that READY alone proves delivery to an external destination.

package/skills/model-lifecycle-management/SKILL.md

@@ -54,3 +54,12 @@ Report immutable input identities, returned run/job/package IDs, status, validat
 Route deployment validation, creation, rollout state, approval, promotion,
 rollback, offline operations, and deployment-bound prediction to the separate
 `model-deployment-management` Skill.
+
+At task completion, suggest only the next gate supported by the observed
+evidence: a valid DatasetArtifact can lead to training, a valid ModelArtifact
+to evaluation, a PASS evaluation to registration and packaging, and a READY
+package with immutable digest to `$model-deployment-management`. If a gate
+failed, suggest inspection or correction of that gate instead. A zero-shot
+prediction is complete on its own; do not imply it produced an artifact or
+requires training. These are optional next actions, not authorization to
+submit or mutate resources.

package/skills/prediction-report-management/SKILL.md

@@ -0,0 +1,80 @@
+---
+name: prediction-report-management
+description: Manage versioned business-owned prediction report destinations, schedules, and run diagnostics through the ITSMP ml-platform CLI. Use for configuring HTTP delivery and token acquisition, creating or changing scheduled reports, pausing or resuming them, or inspecting report runs. Do not use for model deployment, direct prediction, Kubernetes administration, or image releases.
+---
+
+# Prediction Report Management
+
+Use the `ml-platform` CLI installed with this Skill's npm release. Reporting is a
+control-plane workflow: the API runs schedules and sends predictions from a
+READY deployment; model Serving does not receive destination credentials.
+
+## Read-only checks
+
+```bash
+ml-platform --profile server health
+ml-platform --profile server list-destination-profiles
+ml-platform --profile server get-destination-profile NAME v1
+ml-platform --profile server list-prediction-report-schedules
+ml-platform --profile server get-prediction-report-schedule SCHEDULE_ID
+ml-platform --profile server list-prediction-report-runs SCHEDULE_ID
+ml-platform --profile server get-prediction-report-run RUN_ID
+```
+
+Check the deployment with `get-deployment DEPLOYMENT_ID` before creating a
+schedule. The deployment must support `feature_lookup` and have a usable
+inference horizon, supplied by either the schedule request or deployment
+policy. A report run may be `SKIPPED` when the selected revision is not READY.
+
+## Configure a destination
+
+The `PROFILE_JSON` argument is a local JSON file, not inline JSON. Include
+`schema_version`, `name`, `version`, `base_url`, `path`, authentication, and
+response success rules as required by the target. For token-based auth, set
+`authentication.token_request` parameters and `token_path`; configure both
+token and report business success rules when HTTP 2xx alone is insufficient.
+
+```bash
+ml-platform --profile server validate-destination-profile PROFILE_JSON
+ml-platform --profile server create-destination-profile PROFILE_JSON
+ml-platform --profile server get-destination-profile NAME v1
+```
+
+Validation checks schema and local policy; it does **not** call the token or
+report endpoint. Versions are immutable. Changes to URL, auth, token extraction,
+or success rules require a new profile version and a schedule update to point
+to it. `get` and `list` redact credential values; they cannot reconstruct the
+original file. Keep credential-bearing JSON outside the repository, restrict
+file access, and do not print its contents or put secrets in shell arguments,
+logs, examples, or Agent answers. Destination URL and token credentials are
+business configuration submitted through this API, not Kubernetes Secrets.
+
+## Manage a schedule
+
+Creating a schedule starts reporting when `enabled` is true. Obtain explicit
+authorization for the exact destination and deployment before creation, update,
+pause, or resume. Use `enabled: false` for a staged schedule that should not
+run yet. `SCHEDULE_JSON` contains `deployment_id`, five-field `cron`, IANA
+`timezone`, `request.entity`, and `destination: {"name":"...","version":"v1"}`.
+Do not assume that an existing destination profile is bound to any schedule.
+
+```bash
+ml-platform --profile server create-prediction-report-schedule SCHEDULE_JSON
+ml-platform --profile server update-prediction-report-schedule SCHEDULE_ID CHANGES_JSON
+ml-platform --profile server pause-prediction-report-schedule SCHEDULE_ID
+ml-platform --profile server resume-prediction-report-schedule SCHEDULE_ID
+```
+
+Updates accept only `name`, `cron`, `timezone`, `request`, `destination`,
+`max_attempts`, and `timeout_seconds`; pause/resume control `enabled`.
+After mutation, fetch the schedule and its runs. Report the schedule ID,
+revision, destination name/version, next run time, latest run status, and
+redacted error code. Do not claim delivery succeeded based only on validation
+or schedule creation.
+
+At completion, offer a next step only when it fits the observed state: after a
+profile is created, optionally bind it to a schedule for a suitable deployment;
+after a disabled schedule is verified, optionally resume it with explicit
+authorization; after a run, inspect failure evidence or confirm `SUCCEEDED`.
+Do not suggest another configuration step merely to keep the workflow going,
+and do not treat a suggestion as permission to enable external delivery.

package/skills/prediction-report-management/agents/openai.yaml

@@ -0,0 +1,4 @@
+interface:
+  display_name: "定时预测上报管理"
+  short_description: "管理预测上报目标、计划及运行记录"
+  default_prompt: "使用 $prediction-report-management 管理定时预测上报配置、计划和运行记录。"