@thinkingai/ae-cli 1.0.24 → 1.0.27

package/skills/ae-analysis/references/drilldown_users.md

@@ -55,14 +55,14 @@ ae-cli analysis +drilldown_users --dry-run
 | `--include_total` | No | Whether to query users from the total row. Default: false |
 | `--relation_val` | No | Relation value used in relation analysis scenarios |
 | `--use_cache` | No | Whether to use cache. Default: true |
-| `--limit` | No | Optional limit. Default: 20, maximum: 50. |
+| `--limit` | No | Optional limit. Default: 1000, maximum: 100000. |
 | `--offset` | No | Optional offset. Default: 0. |
 | `--timeout_minutes` | No | Query timeout in minutes. If the query exceeds this time, it will be cancelled automatically. |
 
 ## Decision Rules
 - Drilldown positioning parameters must come from upstream analysis results and must not be fabricated.
 - `qp` must not be rewritten or "simplified"; it should remain consistent with the source analysis.
-- For pagination, use `--limit` and `--offset` together. Default limit is 20.
+- For pagination, use `--limit` and `--offset` together. Default limit is 1000, maximum 100000.
 - For the first run, it is recommended to pass only the required parameters (`--project_id`, `--model_type`, `--qp`) and add optional parameters after confirming the chain works.
 - Wrap JSON parameters in single quotes (for example `--qp '{}'`, `--drilldown_groups '{}'`) to avoid shell escaping issues.
 - When dates/time ranges are involved, validate with a short range first and then expand gradually.

package/skills/ae-analysis/references/query_adhoc.md

@@ -95,7 +95,7 @@ ae-cli analysis +query_adhoc --project_id <project_id> --model_type <model_type>
 | `--is_sort_by_columns` | No | Whether to sort query results by columns. Default: false |
 | `--resolve_recent_day` | No | Whether to resolve relative time expressions such as "last 7 days". If omitted, service auto-resolves when `qp.eventView.recentDay` exists and `startTime`/`endTime` is incomplete; otherwise defaults to false. |
 | `--fields` | No | Optional fields to return. Must match column names in result. Invalid fields cause INVALID_FIELDS error. |
-| `--limit` | No | Optional limit. Default: 20, maximum: 50. |
+| `--limit` | No | Optional limit. Default: 1000, maximum: 100000. |
 | `--offset` | No | Optional offset. Default: 0. |
 | `--timeout_minutes` | No | Query timeout in minutes. If the query exceeds this time, it will be cancelled automatically. |
 
@@ -103,7 +103,8 @@ ae-cli analysis +query_adhoc --project_id <project_id> --model_type <model_type>
 - On the first run, start with only the required parameters (`--project_id`,`--model_type`,`--qp`), and add optional parameters after confirming the path works.
 - Do not call this command with placeholder QP such as `{}`. For builder-supported models, wait for builder `status=generated`; for non-builder models, build QP from verified schema/metadata first.
 - For builder-supported models, do not run metadata/schema lookup to "help" the builder. The builder is the metadata resolver.
-- For pagination, use `--limit` and `--offset` together. Default limit is 20.
+- For pagination, use `--limit` and `--offset` together. Default limit is 1000, maximum 100000.
+- For `model_type=sql`, use exactly one row limit: either put `LIMIT` in the SQL, or use `--limit`. If both are present they must be equal — `query_adhoc` rejects mismatched values to avoid silently sampled or duplicated results. When the SQL already has a trailing `LIMIT`, you may omit `--limit` and the SQL governs.
 - Use `--fields` to select specific columns for lighter response payloads.
 - `event`, `retention`, `funnel`, and `prop_analysis` must not manually craft QP with schema-first flow; call the builder and use the returned `qp`.
 - For non-builder models, `qp` must satisfy both schema structure and project metadata constraints.

package/skills/ae-analysis/references/query_entity_details.md

@@ -39,7 +39,7 @@ ae-cli analysis +query_entity_details --dry-run
 | `--properties` | No | Optional display properties JSON, for example `[{"columnName":"#user_id","tableType":"0"},{"columnName":"device_id","tableType":"0"}]`. If provided, property names should come from `analysis_meta +list_properties` in the same `project_id`. |
 | `--sort_by` | No | Optional sort field                                                                                                                                                                                                                     |
 | `--sort_order` | No | Optional sort order. Supported values: asc and desc                                                                                                                                                                                     |
-| `--limit` | No | Optional result limit. Default: 20, maximum: 50                                                                                                                                                                                         |
+| `--limit` | No | Optional result limit. Default: 1000, maximum: 100000                                                                                                                                                                                         |
 | `--zone_offset` | No | Time zone offset. For example, UTC+8 is 8                                                                                                                                                                                               |
 | `--use_cache` | No | Whether to use cache. Default: true                                                                                                                                                                                                     |
 

package/skills/ae-engage/SKILL.md

@@ -79,6 +79,12 @@ ae-cli engage +config_channel_list --project_id 1
 # Query the task list
 ae-cli engage +task_list --project_id 1 --req '{"pageNum":1,"pageSize":20}'
 
+# Build a save_task guide before composing the final req
+ae-cli engage +build_task_save_guide --project_id 1 --req '{"context":{"triggerType":2,"channelId":"channel_123"}}'
+
+# Save a task draft (create when req.taskId is omitted)
+ae-cli engage +save_task --project_id 1 --req '{"baseInfo":{"taskName":"Demo Task"},"channelConfig":{"channelType":1,"channelId":"channel_123","groupContentList":[{"contentList":[{"pushLanguageCode":"default","content":"[]"}]}]},"targetConfig":{"targetClusterType":3},"triggerConfig":{"triggerType":2},"controlConfig":{"completionIndicatorDef":{"completionIndicators":[]}}}'
+
 # Query task details
 ae-cli engage +task_detail --project_id 1 --task_id task_123
 
@@ -182,6 +188,8 @@ For more detailed generation rules, consult these references first:
 ```bash
 ae-cli --dry-run engage +channel_list --project_id 1
 ae-cli --dry-run engage +task_list --project_id 1 --req '{"pageNum":1,"pageSize":20}'
+ae-cli --dry-run engage +build_task_save_guide --project_id 1 --req '{}'
+ae-cli --dry-run engage +save_task --project_id 1 --req '{"baseInfo":{"taskName":"Demo Task"},"channelConfig":{"channelType":1,"channelId":"channel_123","groupContentList":[{"contentList":[{"pushLanguageCode":"default","content":"[]"}]}]},"targetConfig":{"targetClusterType":3},"triggerConfig":{"triggerType":2},"controlConfig":{"completionIndicatorDef":{"completionIndicators":[]}}}'
 ae-cli --dry-run engage +flow_list --project_id 1
 ```
 
@@ -190,6 +198,8 @@ ae-cli --dry-run engage +flow_list --project_id 1
 More detailed single-command guidance is available in the business-oriented `references/` directory:
 
 - `references/channel-list.md`
+- `references/build-task-save-guide.md`
+- `references/save-task.md`
 - `references/task-list.md`
 - `references/config-item-list.md`
 - `references/flow-list.md`
@@ -204,7 +214,7 @@ This split documentation structure is easier to extend later, because commands w
 
 ### task
 
-`+task_data_overview`, `+task_data_detail`, `+task_metric_detail`, `+task_experiment_report`, `+task_detail`, `+task_list`, `+task_stats`, `+manage_task`
+`+task_data_overview`, `+task_data_detail`, `+task_metric_detail`, `+task_experiment_report`, `+task_detail`, `+task_list`, `+task_stats`, `+build_task_save_guide`, `+save_task`, `+manage_task`
 
 ### config
 
@@ -225,4 +235,17 @@ The following commands are write operations. Confirm that the user intent is exp
 - Channels and config channels: `+add_channel`, `+delete_channel`, `+update_channel_status`, `+delete_config_channel`, `+update_config_channel_status`
 - Strategies and config items: `+delete_config_item`, `+copy_config_template`, `+manage_strategy`
 - Flows: `+save_flow`, `+modify_flow_base_info`, `+manage_flow`, `+delete_flow`
-- Tasks: `+manage_task`
+- Tasks: `+save_task`, `+manage_task`
+
+For task draft creation or update, use this workflow:
+
+1. `ae-cli engage +channel_list --project_id <projectId>`
+2. `ae-cli engage +build_task_save_guide --project_id <projectId> --req '{...}'`
+3. If the guide says QP-derived fields are needed, call `ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition`
+4. `ae-cli engage +save_task --project_id <projectId> --req '{...}'`
+
+`+build_task_save_guide` is a read-only helper. It returns scenario-specific required fields, channel content schema, unsupported combinations, examples, and a handoff template for `save_task`.
+
+`+save_task` only saves a draft. It does not submit approval, does not start sending, and does not trigger task execution. If `req.taskId` is omitted it creates a new draft; if `req.taskId` is present it updates an existing draft. Update mode only supports draft tasks, and omitted fields can inherit from the existing draft before validation.
+
+The audience schema query `ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition` is not a fixed preflight step. Call it only when the guide indicates that you must construct `targetConfig.qp`, `triggerConfig.triggerRule`, `clientConfig.clientQp`, or `completionIndicatorDef.event`.

package/skills/ae-engage/references/build-task-save-guide.md

@@ -0,0 +1,283 @@
+# ae-engage `+build_task_save_guide`
+
+Build a scenario-specific guide for constructing `save_task.req`.
+
+Mapped command: `ae-cli engage +build_task_save_guide`
+
+This is a read-only helper. It does not save tasks, submit approvals, or trigger execution. Its job is to tell you how to build a valid `save_task` request for the current scenario.
+
+---
+
+## 1. Recommended Workflow
+
+Use this sequence when creating or updating a task draft:
+
+1. Query channels with `ae-cli engage +channel_list --project_id <projectId>`.
+2. Call `ae-cli engage +build_task_save_guide --project_id <projectId> --req '{...}'`.
+3. If the guide says QP-derived fields are required, call:
+
+```bash
+ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition
+```
+
+4. Build the final grouped `save_task.req`.
+5. Submit with `ae-cli engage +save_task --project_id <projectId> --req '{...}'`.
+
+Important:
+
+- Do not treat the audience schema query as a fixed preflight step.
+- Call it only when the guide indicates that you need to construct:
+  - `targetConfig.qp`
+  - `triggerConfig.triggerRule`
+  - `clientConfig.clientQp`
+  - `completionIndicatorDef.event`
+
+---
+
+## 2. CLI Shape
+
+The CLI call shape is:
+
+```bash
+ae-cli engage +build_task_save_guide --project_id <projectId> --req '<req-json>'
+```
+
+This CLI requires `--req`, but `{}` is valid when you want a generic guide without scenario context.
+
+Common request patterns:
+
+```bash
+ae-cli engage +build_task_save_guide --project_id 1 --req '{}'
+ae-cli engage +build_task_save_guide --project_id 1 --req '{"context":{"triggerType":2,"channelId":"channel_123"}}'
+ae-cli engage +build_task_save_guide --project_id 1 --req '{"draft":{"baseInfo":{"taskName":"Demo Task"}}}'
+```
+
+Supported request fields:
+
+- `context`
+  - a lightweight scenario selector
+  - use it when you know key dimensions such as trigger type, audience type, channel, or experiment mode
+- `draft`
+  - a partial `save_task` request
+  - use it when you already have part of the final payload and want missing fields, examples, or corrections
+- `reqDraft`
+  - alias of `draft`
+- `detailLevel`
+  - `brief`: core requirements only
+  - `full`: includes enums, unsupported cases, examples, and self-check information
+
+---
+
+## 3. `context` and `draft`
+
+### 3.1 `context`
+
+Use `context` when you want scenario-specific guidance without writing a partial `save_task.req`.
+
+Typical fields:
+
+- `triggerType`
+- `targetClusterType`
+- `channelType`
+- `channelId`
+- `channelTemplateId`
+- `enableExp`
+- `expIndicatorBizType`
+
+Example:
+
+```json
+{
+  "context": {
+    "triggerType": 2,
+    "targetClusterType": 2,
+    "channelId": "channel_123"
+  }
+}
+```
+
+### 3.2 `draft`
+
+Use `draft` when you already have a partial `save_task.req` and want the guide to tell you:
+
+- what is still missing
+- what is invalid
+- whether the payload is already close to submit-ready
+- what template or schema to use to finish it
+
+Example:
+
+```json
+{
+  "draft": {
+    "baseInfo": {
+      "taskName": "Demo Task"
+    },
+    "triggerConfig": {
+      "triggerType": 2
+    }
+  }
+}
+```
+
+---
+
+## 4. Output Structure
+
+The guide returns structured metadata rather than free-form prose. The top-level sections are:
+
+- `valid`
+- `errors`
+- `warnings`
+- `missingInputs`
+- `saveTaskContract`
+- `scenario`
+- `requiredInput`
+- `fieldRules`
+- `handoff`
+
+### 4.1 `valid`
+
+- `true` means the current scenario or draft has no structural blocking errors
+- `false` means you should inspect `errors`, `missingInputs`, and `handoff.blockingPlaceholders`
+
+### 4.2 `errors`
+
+Use `errors` for hard blockers such as:
+
+- invalid enum values
+- unsupported combinations
+- unsupported trigger types
+
+The guide may expose invalid values separately so you can correct them instead of silently dropping them.
+
+### 4.3 `warnings`
+
+Use `warnings` for non-fatal guidance, for example:
+
+- channel content schema is unavailable because no `channelId` was provided
+- click-rate experiment setup still needs a channel context
+
+### 4.4 `missingInputs`
+
+Use `missingInputs` as the to-do list for fields you still need before you can submit `save_task`.
+
+Typical examples:
+
+- missing `baseInfo.taskName`
+- missing `channelConfig.channelId`
+- missing `channelConfig.groupContentList`
+- missing `triggerConfig.triggerTime` for scheduled single tasks
+
+### 4.5 `saveTaskContract`
+
+This section describes the high-level contract:
+
+- final tool is `save_task`
+- required preflight is `query_channel_list -> build_task_save_guide`
+- QP-derived fields may require `ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition`
+- `save_task.req` must be a grouped JSON object
+
+### 4.6 `scenario`
+
+This section shows the resolved scenario, such as:
+
+- create vs update draft
+- trigger type
+- audience type
+- channel type / channel id
+- whether the scenario is unsupported
+
+### 4.7 `requiredInput`
+
+This section tells you which grouped blocks and fields are required in the current scenario.
+
+Use it to understand what must appear in:
+
+- `baseInfo`
+- `channelConfig`
+- `targetConfig`
+- `triggerConfig`
+- `controlConfig`
+- `expConfig` when experiment mode is enabled
+
+### 4.8 `fieldRules`
+
+This is the most important construction section.
+
+It includes:
+
+- grouped block rules
+- structured conditional rules
+- related-parameter rules
+- unsupported fields / values / combinations
+- `channelContentSchema`
+- enum values and wrong examples when `detailLevel=full`
+
+#### `fieldRules.channelContentSchema`
+
+Treat this as the source of truth for channel content construction.
+
+Use it to get:
+
+- valid keys
+- expected item shape
+- channel-specific examples
+- params that must be copied exactly, such as `paramType`
+
+Do not invent free-form content items such as:
+
+```json
+[{"text":"hello"}]
+```
+
+Instead, use the valid item structure and put message text into `value`.
+
+### 4.9 `handoff`
+
+This is the final section before `save_task`.
+
+Important fields:
+
+- `reqTemplate`
+  - a scenario-aware grouped request template
+  - use it as a starting point, not as unquestioned final truth
+- `readyToSubmit`
+  - `true` means the current scenario or draft has no blocking placeholders
+- `blockingPlaceholders`
+  - unresolved placeholders that still block submission
+- `selfCheckList`
+  - a final checklist before calling `save_task`
+- `successUrlRule`
+  - after `save_task` succeeds, build the task detail URL from current page origin plus the relative hash
+  - do not hardcode host, IP, or port
+
+---
+
+## 5. How to Use the Guide in Practice
+
+Recommended usage pattern:
+
+1. resolve a real `channelId` with `+channel_list`
+2. call `+build_task_save_guide`
+3. read `fieldRules.channelContentSchema`
+4. read `handoff.reqTemplate`
+5. fix everything in `blockingPlaceholders`
+6. if the guide points to QP-derived fields, call:
+
+```bash
+ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition
+```
+
+7. fill those fields using the real condition schema
+8. call `+save_task`
+
+---
+
+## 6. Safety Notes
+
+- `+build_task_save_guide` is read-only; it does not save a draft
+- do not skip the guide and build `save_task.req` from memory alone
+- do not treat the audience schema query as mandatory for every task
+- do not invent `channelId`, `clusterKey`, content keys, or QP structures
+- if the guide exposes an unsupported scenario, correct it before calling `save_task`

package/skills/ae-engage/references/save-task.md

@@ -0,0 +1,304 @@
+# ae-engage `+save_task`
+
+Create or update a Hermes task draft.
+
+Mapped command: `ae-cli engage +save_task`
+
+This command is the final write step. Do not use it as the first step in task construction.
+
+Recommended workflow:
+
+1. `ae-cli engage +channel_list --project_id <projectId>`
+2. `ae-cli engage +build_task_save_guide --project_id <projectId> --req '{...}'`
+3. If the guide indicates QP-derived fields are needed, call:
+
+```bash
+ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition
+```
+
+4. Build the final grouped `req`
+5. `ae-cli engage +save_task --project_id <projectId> --req '{...}'`
+
+The audience schema query is not a fixed preflight step. Call it only when the guide indicates that you must construct:
+
+- `targetConfig.qp`
+- `triggerConfig.triggerRule`
+- `clientConfig.clientQp`
+- `completionIndicatorDef.event`
+
+For the full guide contract, request format, return sections, and handoff usage, read:
+
+- `references/build-task-save-guide.md`
+
+---
+
+## 1. General Principles
+
+`+save_task` accepts only the final grouped draft-save payload that Hermes can validate and persist.
+
+The command supports two modes:
+
+- create mode: omit `req.taskId`
+- update mode: include `req.taskId` for an existing draft task
+
+Regardless of mode, this tool only saves a draft:
+
+- it does not submit approval
+- it does not start sending
+- it does not register or start trigger execution
+
+The CLI call shape is:
+
+```bash
+ae-cli engage +save_task --project_id <projectId> --req '<req-json>'
+```
+
+Notes:
+
+- `projectId` is injected into top-level `req` by the CLI
+- if the caller also passes `req.projectId`, the outer `--project_id` wins
+- the whole `req` must be a JSON object, not a stringified JSON string
+- for update mode, Hermes only allows modifying draft tasks
+- in update mode, omitted fields are backfilled from the existing draft before validation, so partial draft updates are allowed
+
+---
+
+## 2. Required Workflow
+
+### 2.1 Query Real Channels First
+
+Run:
+
+```bash
+ae-cli engage +channel_list --project_id <projectId>
+```
+
+Purpose:
+
+- get the real project channel list
+- resolve the real `channelId`
+- confirm the final `channelType`
+
+Never invent a `channelId`.
+
+### 2.2 Build the Scenario Guide
+
+Run:
+
+```bash
+ae-cli engage +build_task_save_guide --project_id <projectId> --req '{...}'
+```
+
+Use the guide to determine:
+
+- required grouped blocks
+- required fields for the current trigger and audience mode
+- unsupported combinations
+- correct content schema
+- current handoff template
+- whether the current payload is already close to submit-ready
+
+### 2.3 Query Audience Condition Schema Only When Needed
+
+If the guide indicates that QP-derived fields are needed, run:
+
+```bash
+ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition
+```
+
+Use the returned schema to build:
+
+- `targetConfig.qp`
+- `triggerConfig.triggerRule`
+- `clientConfig.clientQp`
+- `completionIndicatorDef.event`
+
+Do not call this command by default for every task. It is conditional, not mandatory.
+
+### 2.4 Build Final `req` and Save
+
+Only after the above steps should you construct the final grouped `req` and submit it with `+save_task`.
+
+---
+
+## 3. Final `req` Shape
+
+The final object passed to `--req` should have this grouped shape:
+
+```json
+{
+  "taskId": "<string, optional>",
+  "baseInfo": {},
+  "channelConfig": {},
+  "targetConfig": {},
+  "triggerConfig": {},
+  "controlConfig": {},
+  "expConfig": {},
+  "activityConfig": {},
+  "clientConfig": {}
+}
+```
+
+Top-level notes:
+
+- `taskId` is used only in update mode
+- `baseInfo` / `channelConfig` / `targetConfig` / `triggerConfig` / `controlConfig` are the main blocks
+- optional blocks should usually be omitted rather than filled with invented values
+- use `build_task_save_guide` output as the source of truth for the current scenario
+
+---
+
+## 4. How To Build Major Blocks
+
+### 4.1 `taskId`
+
+Use `taskId` only when you are updating an existing draft task.
+
+- create mode: omit `taskId`
+- update mode: include `taskId`
+- update mode can send only the fields that need changing
+- update mode still fails if the referenced task is not in draft status
+
+### 4.2 `channelConfig`
+
+Core fields:
+
+- `channelType`
+- `channelId`
+- `groupContentList`
+- optional `channelTemplateId`
+
+Rules:
+
+- derive `channelType` and `channelId` from real channel metadata
+- do not guess template IDs
+- `groupContentList` maximum size is `5`
+- `occasionKeys` are parsed from content automatically and are not accepted as input
+
+Content guidance:
+
+- do not invent channel-specific payload structures from memory
+- use `fieldRules.channelContentSchema` from `build_task_save_guide`
+- take valid keys, expected item shape, and examples from the guide
+- when schema fields include `paramType`, copy it exactly
+- do not use free-form content items such as `{"text":"..."}` as the primary pattern
+
+### 4.3 `targetConfig`
+
+Use the guide to decide which audience shape applies:
+
+- `targetClusterType=1`: custom audience, requires `qp`
+- `targetClusterType=2`: existing cluster, requires `clusterKey`
+- `targetClusterType=3`: all users, forbids both `qp` and `clusterKey`
+
+If `qp` is required, construct it from:
+
+```bash
+ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition
+```
+
+### 4.4 `triggerConfig`
+
+Use the guide to determine the current trigger rule set:
+
+- `triggerType=0`: requires `triggerTime`
+- `triggerType=1`: requires `startDate`, `endDate`, `triggerCrontab`
+- `triggerType=2`: manual, no `triggerRule`
+- `triggerType=3/4/5`: requires `triggerRule`
+
+Rules:
+
+- `triggerType=6` is not supported
+- use the guide for cron format and wrong-example checks
+- if `triggerRule` is needed, build it from:
+
+```bash
+ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition
+```
+
+### 4.5 `controlConfig`
+
+Minimum required field:
+
+- `completionIndicatorDef`
+
+When the guide points to event-based completion or experiment-driven main-goal rules, use:
+
+```bash
+ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition
+```
+
+to build `completionIndicatorDef.event`.
+
+Important constraints that still apply:
+
+- `doNotDisturb.enableDoNotDisturb=true` requires `startTime` and `endTime` in `HH:mm`
+- `pushDelay.enablePushDelay=true` requires valid `delayType`, and its dependent fields must match the selected mode
+- `pushDelay.delayType=1` is only valid for `client_push`
+- `pushDelay.delayUnit` must stay within `week`, `day`, `hour`, `minute`, `second`
+- `timeoutControl.enableTimeoutControl=true` requires valid `time` and `unit`
+- `timeoutControl.unit` must stay within `day`, `hour`, `minute`
+
+### 4.6 `clientConfig`
+
+Use this block only when client-side conditions are needed.
+
+If `clientConfig.clientQp` is required, build it from:
+
+```bash
+ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition
+```
+
+Only call that schema query when the guide explicitly shows that the field is needed.
+
+---
+
+## 5. Final Self-Check Before `+save_task`
+
+Before submission, verify:
+
+1. `channelId` comes from a real channel query.
+2. `build_task_save_guide` has already been called for the current scenario or partial draft.
+3. `fieldRules.channelContentSchema` was used as the source of truth for content structure.
+4. Any required QP-derived fields were built from `ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition`.
+5. `targetClusterType` matches the presence or absence of `clusterKey` / `qp`.
+6. `triggerType` matches the provided scheduling or event fields.
+7. `completionIndicatorDef` is present and structurally valid for the current scenario.
+8. `taskId` is omitted for create mode and present only for updating a draft.
+9. No unsupported `triggerType=6` is used.
+10. No placeholder IDs or fabricated resource names remain in the request.
+
+---
+
+## 6. Standard Example Flow
+
+Use this style of workflow, rather than jumping directly to `+save_task`:
+
+```bash
+ae-cli engage +channel_list --project_id 1
+ae-cli engage +build_task_save_guide --project_id 1 --req '{"context":{"triggerType":2,"channelId":"channel_123"}}'
+ae-cli engage +save_task --project_id 1 --req '{...final grouped req...}'
+```
+
+If the guide indicates QP-derived fields are needed, insert:
+
+```bash
+ae-cli analysis_audience +get_cluster_definition_schema --cluster_type condition
+```
+
+before building the final `req`.
+
+---
+
+## 7. Safety Constraints
+
+This command is a write operation.
+
+- Do not treat `save_task` as “submit and launch task”
+- Do not bypass `build_task_save_guide`
+- Do not pass the whole `req` as a JSON string
+- Do not use `triggerType=6`
+- Do not invent `channelId`, `clusterKey`, audience definitions, or content keys
+- Do not use `taskId` for a non-draft task
+- Do not pass `occasionKeys`; Hermes derives them from content
+- Do not call the audience schema query as a reflex; call it only when guide output says QP-derived fields are needed