yetong-timer

Ye Tong Daily Trending Topic Selection

What is fixed is the retrieval, filtering, and output rules; the city, role, target members, business, speakable perspectives, and expression come from this coach's own DNA. Deliver 3 specific shootable questions each day, or the creation result of a real native task. Douyin is a source of trending topics, not a reason to change the audience into fitness peers; topics can be used for Xiaohongshu or real-person short videos, but cross-platform popularity is not guaranteed.

First determine the action

Create: "Based on my DNA help me do scheduled task topic selection" and "give me trending topics on a daily schedule" enter creation preparation. Only ask follow-up questions about the execution time, repeat frequency, time zone, and material location that are still missing; do not ask again about information the user has already made clear. If there is no time, do not arbitrarily set 9 a.m.; if there is no time zone, do not infer it from IP or city.
Trial run: "find 3 trending topics today" and "do a trial run first, do not create" execute the content task only once and do not call scheduling writes.
Run: when started from a registered native task and binding information already exists, execute only this topic selection. A run does not create itself again, nor does it arbitrarily change the frequency.
A request to create/modify this Skill is not a request to create a background task; installing this Skill also does not automatically establish a task. An ordinary "give 5 topics" belongs to yetong-topic; "what promotion to post in Moments every day for this course" belongs to yetong-promo, not scheduled authorization. Arbitrary reminders, scheduled script writing for course marketing, or performance management are not temporarily handled by this module.

Personal profile threshold

Only read the user's own person-dna.md path/private document reference that the user explicitly provides or that is already bound to the current task. If the location is missing, only ask for the location; do not scan private directories, and do not substitute examples, other people's materials in host memory, or Ye Tong's own profile. When there are multiple people's profiles or an identity conflict, first confirm the subject.
When an executable file environment is available, run from this Skill directory python3 scripts/check_profile.py <档案路径>. The exit code is 0 and after the output starts with active: , each fact and permission must still be checked item by item. When there is a readable cloud document but no Python, the same schema, status, confirmation date, and six required sections can be checked item by item; clearly state that this is a manual structural verification and do not claim the script was run. Do not accept a verbal claim of only "I have activated it."
When it cannot be read, is not activated, or the structure is invalid, display the following prompt and stop personalized topic selection and new task creation; if only the reference location is missing, first ask for the location. Do not activate it yourself, and do not provide a "generic scheduled task" that borrows private materials.

The persona DNA is not yet activated. Currently only generic methods can be called, and personalized topic selection, real-experience mapping, and personal expression calibration cannot be enabled. Only after completing the "Persona Foundation Interview" and confirming the profile can the full Ye Tong content method engine be unlocked.

The person's own role, service area, local audience, real concerns, related business, and publicly shareable experiences/service actions are the basis. When the profile is activated but current key audience/city materials are missing, only ask about that gap; the old version is valid schema_version: 2 Full re-visiting is not required. Pain points must not be inferred from labels such as women or mothers. For each permission rule, read the in-package yetong-topic/references/profile-contract.mdfirst; when it cannot be retrieved, follow the permission boundaries in this section and in the hotspot rules.

Topic selection rules and dependencies

On every run, re-read the currently bound profile and the hotspot retrieval, filtering, and output rules, and do not reuse expired retrieval results.

Prefer reading the installed yetong-topic/SKILL.md and yetong-topic/references/topic-method.md, and reuse the eight major elements and four types of content methods therein. This module's fixed quantity, hotspot basis, and output format take priority over the conventional topic selection defaults of 8 topics, a full candidate table, or a consultation orientation.
If the current task is bound to verifiable eight-major-elements Skill/project rules, first check whether it is this set of methods and the current version, then read it; a similar name must not be treated as having been read.
When the real method resources cannot be retrieved, use the embedded eight major elements and four types of content rules in the hotspot rules, write "Rule status: executed per embedded rules," and explain that the method resources were not obtained. Do not claim to have called a missing Skill, and do not automatically install dependencies.
The fixed goal is to build professional trust with your local potential members and increase relevant exposure. It is allowed to enter from cross-industry hotspots and obtain non-local views, but each topic must be able to answer "Which type of my target members, in what situation, can understand or solve what specific problem?" and explain "Why this is suitable for me, this coach, to talk about." Do not write it as a marketing lesson for fellow coaches; do not require every topic to sell a course, include a purchase code, or drive conversion.

Execution

Trial run/run: profile passes → conduct an online search for that run and verify time and data → form 6–9 candidates within the range actually obtained → filter down to 3 different problems prioritizing audience business relevance → deliver only the final cards. When verification is impossible or candidates are insufficient, explicitly downgrade per reference; do not fabricate candidates, hot lists, or metrics.

Create: first read native scheduling and private material binding. Register the task only when this creation authorization, the complete time configuration, usable native scheduling capability, and your own profile and methods that the future execution environment can actually read are all satisfied. Prefer a trial run first to check whether the content and references are usable; a trial run cannot replace successful scheduling. When the host has no native capability, deliver the copyable task instructions and the still-missing conditions, explicitly state "not created," and do not write system cron, set up a resident background process, or borrow tools from another platform.

Output and stopping

A single piece of content outputs only the final 3 topics and the necessary basis per the hotspot rules, does not show the full candidate pool, does not expand the body, opening, script, storyboard, or cover, and does not publish or contact clients.
Only after native scheduling succeeds and a verifiable task identifier, enabled status, time/frequency, and time zone are obtained may creation success be confirmed, and return the task name, identifier, enabled status, execution time zone, repeat frequency, next run time, and verification basis. When the above creation evidence is missing, write "creation result unconfirmed" and do not create repeatedly. If the next time is not obtained, write "not obtained"; do not treat an estimate as a platform return value; do not claim that on-time execution has been accepted.
When unsuccessful, explain the reason, which preparations have been completed, and which configuration is missing; a task card and prompt do not equal having been created. When the result is unknown, check the native list first and do not create repeatedly.
Review the active and per-item public permissions on each run; when materials become invalid, stop that personalized output and prompt for a fix, without changing the profile on your own or pausing/deleting the native task. Private DNA, task configurations, historical topic selections, member profiles, and test records do not enter the public Skill directory or GitHub.