Watch a ThinkingAI agent run the full loop,
from data collection to closed-loop action.

Please use the ae-generate-tracking-plan Skill to scan the current project code and generate and upload a tracking plan that follows AE conventions.
[Target AE project]
AE Host: {{AE_HOST}}
Project ID: {{AE_PROJECT_ID}}
The AE CLI is already authorized. Use the Host and Project ID above directly. Do not ask me to log in again or look up the project ID.
[Business context]
Product background: Identify the product form, target users, and core business from the current project code. Do not ask me to fill this in again.
Required scenarios: Identify the core features and real business flows from the current project code. Cover only scenarios that actually exist. Do not copy the examples or invent features.
[Execution requirements]
›Use the current project code as the primary source. Follow the Skill to generate, confirm, validate, archive, and upload the plan.
›Name events in snake_case. Note the trigger timing and required properties for each event. Every event, event property, common event property, and user property must have a human-readable display name filled in. None may be left blank.
›Show me the event table and property table for confirmation first. Do not upload before I confirm.
›After confirmation, use ae-cli to generate and validate the draft/xlsx, then upload to the Project ID above. After upload, run ae-cli tracking plan get to read it back and confirm.
›After a successful upload, run ae-cli tracking plan sync-display-names --project-id {{AE_PROJECT_ID}} --draft <path to the draft.json you generated>. Fill in only the blank display names. Do not overwrite display names that already exist in AE.
›If it returns missing_in_metadata, list the events or properties that have not appeared yet and retry after the data is reported. Do not treat this as a failed plan upload.
›Finally, return the AE tracking-plan page, the local archive file path, and the display-name sync result.

Please use the ae-generate-tracking-code Skill to read the current project code and the .ae-tracking/draft.json generated in the previous step, and complete the tracking-code integration.
[AE integration info]
AE Host: {{AE_HOST}}
Project ID: {{AE_PROJECT_ID}}
SERVER_URL: {{AE_SERVER_URL}}
APP_ID: {{AE_APP_ID}}
[Configuration rules]
›SERVER_URL is already derived from the current AE address: replace only the -web- in the domain with -receiver-, and keep the protocol, the rest of the domain, and the port.
›Project ID and APP_ID are two different fields. Do not mix them up.
›When SERVER_URL and APP_ID are not UNAVAILABLE, they have already been provided and confirmed by the current AE project. Write them directly into meta.server_url and meta.app_id in .ae-tracking/draft.json. Do not ask again.
›If APP_ID is UNAVAILABLE, query it first through ae-cli using the Project ID above. If it cannot be found, prompt me to go to AE under Project Management > Integration Config and copy the APP_ID.
›If SERVER_URL is UNAVAILABLE, do not substitute the AE Host. Prompt me to get the public data-reporting address from AE under Project Management > Integration Config.
[Code change requirements]
›Read .ae-tracking/draft.json first. If it does not exist, use the AE Host and Project ID above to fetch the uploaded tracking plan.
›Analyze the current project code first and list the files and tracking locations you plan to add or change.
›Provide a change preview first. Do not change any code before I confirm.
›After confirmation, complete the SDK configuration and the tracking-code changes.
›When done, run the project build and the related checks.

Please continue using the ae-generate-tracking-code Skill to generate a debug validation script for the current project, and guide me through verifying the data reporting.
[Verification target]
AE Host: {{AE_HOST}}
Project ID: {{AE_PROJECT_ID}}
Platform or SDK: {{e.g. Web JavaScript SDK, Android SDK}}
[Verification requirements]
›Choose representative business events from the tracking plan and generate a validation script.
›Generate a stable debug device ID. First run ae-cli tracking debug-device list to see existing devices.
›If the device does not exist, create it with ae-cli tracking debug-device add, then apply it with ae-cli tracking debug-device select.
›After running the validation script, use ae-cli tracking debug-data list to query the last hour of debug data. Filter by event name if needed.
›Only judge the verification successful when has_data is true and the event names, property structures, and error-field checks all pass.
›After the debug data is confirmed to arrive, run ae-cli tracking plan sync-display-names --project-id {{AE_PROJECT_ID}} --draft <path to the draft.json you generated> to fill in the blank display names for this batch of events and properties. Do not overwrite existing display names.
›Verify locally by default. Only when the CLI capability is unavailable, tell me how to troubleshoot on the AE Debug page.
›If verification fails, explain the cause and how to fix it.
[Additional notes]
{{Other verification requirements; delete if none}}

Please use the ae-kb Skill to build a metrics knowledge base for the current AE project through ae-cli.
[Target project]
AE Host: {{AE_HOST}}
Project ID: {{AE_PROJECT_ID}}
[Goal]
Create a company knowledge base named "marketing-site metrics library". Add the project's uploaded tracking plan and the metric definition docs, compile it, and confirm the wiki is searchable.
[Execution requirements]
›Use ae-cli kb commands only (+create, +upload, +compile, +query).
›After compiling, run one test query: ask for the numerator and denominator of the signup conversion rate, and require the answer to cite its source page.
›Do not invent metric definitions. If a definition is missing from the docs, list it as an open question instead.
›When done, return the knowledge base name, the page count, and the cited answer.

Please use the ae-analysis Skill to create dashboard reports in the current AE project through ae-cli.
[Target project]
AE Host: {{AE_HOST}}
Project ID: {{AE_PROJECT_ID}}
[Analysis goal]
Identify the core business goals and complete the analysis using the current project code, the uploaded tracking plan, and the existing data.
[Example scenarios]
›A 7-day trend of the core events
›A conversion funnel for the core business path
›Next-day retention of new users
[Execution requirements]
›Use the Host and Project ID above and read the project's uploaded tracking plan.
›Do not guess or substitute the project_id, event names, or property names. Use only the events and properties that actually exist in the tracking plan.
›Understand the real business scenarios from the current project code and choose analysis goals that fit this project. The scenarios above are for reference only.
›Create at least 2 analysis reports and one dashboard that matches the project and the analysis goals, and add the created reports to the dashboard.
›Verify that each report queries correctly. If the current data is insufficient, say so clearly. Do not fabricate data.
›When done, return the dashboard link and the report list.

Please use the ae-engage Skill to build a win-back flow in the current AE project through ae-cli.
[Target project]
AE Host: {{AE_HOST}}
Project ID: {{AE_PROJECT_ID}}
[Goal]
Segment last-7-day visitors who typed into the inquiry form but never submitted, and win them back.
[Execution requirements]
›Create two cohorts: high-intent churn (form input in the last 7 days) and light churn (exposure only).
›Draft the message for each branch and stop for my approval. Do not launch anything before I confirm.
›After I confirm, create one canvas with two branches on a daily 09:00 schedule.
›Add a report to the existing dashboard tracking signup conversion of the win-back cohort.
›When done, return the flow name, the node count, and the report link.