yetong-timer

Original：🇨🇳 Chinese

Translated

1 scripts

Based on the local fitness coach's own activated DNA, retrieve verified trending topics from Douyin and trusted public web pages, then screen three topics suited to the coach's target members on each run. Use this when the user says “Help me create a scheduled topic-selection task based on my DNA,” requests daily trending-topic ideas, or asks for a one-off trial run. It can create scheduled tasks through the host's native capabilities, but does not handle arbitrary task scheduling, complete scriptwriting, auto-publishing, or teaching peers how to acquire customers.

4installs

Sourceyetongzhu463-stack/yetong-fitness-xhs

Added on2026-10-15

NPX Install

npx skill4agent add yetongzhu463-stack/yetong-fitness-xhs yetong-timer

SKILL.md Content (Chinese)

View Translation Comparison →

Yetong Daily Trending-Topic Ideas

The retrieval, screening, and output rules are fixed; the city, role, target members, business, angles available to discuss, and expression come from the current coach's own DNA. Each run delivers three specific, shootable questions, or the actual creation result for a native task. Douyin is a source of trends, not a reason to change the audience to other fitness professionals; the topics can be used for Xiaohongshu or live-action short videos, but cross-platform popularity is not guaranteed.

Determine the Action First

Create: Requests such as “Help me create a scheduled topic-selection task based on my DNA” or “Give me trending-topic ideas every day” enter creation preparation. Ask only for missing execution time, repeat frequency, time zone, and location of the source materials; do not ask for information the user has already provided. If no time is given, do not default to 9:00 a.m.; if no time zone is given, do not infer it from the IP address or city.
Trial run: “Find three trending topics today” or “Run a trial first; don't create anything” performs only one content task and does not invoke a scheduling write.
Run: When launched from a registered native task with binding information, execute only the current topic selection. A run does not create another instance of itself or change the recurrence frequency.
Requests to create or modify this Skill are not requests to create a background task; installing this Skill does not automatically create one. An ordinary request such as “Give me five topic ideas” belongs to
```
yetong-topic
```
; “What should I post on WeChat Moments every day to promote this course?” belongs to
```
yetong-promo
```
and does not constitute authorization to schedule anything. This module does not take over arbitrary reminders, scheduled copywriting for course marketing, or performance management on an ad hoc basis.

Requirements for the Coach's Own Profile

Only read the coach's own
```
person-dna.md
```
path or private-document reference that the user explicitly provides or that is already bound to the current task. If the location is missing, ask for it; do not scan private directories, use examples, access other people's information in host memory, or substitute Yetong's own profile. If there are multiple profiles or conflicting identities, confirm the target first.
When the runtime can access executable files, run
```
python3 scripts/check_profile.py <profile path>
```
from this Skill directory. Even when the exit code is
```
0
```
and the output begins with
```
active:
```
, every fact and permission must still be checked. When a cloud document is readable but Python is unavailable, manually verify the same schema, status, confirmation date, and all six required sections. Explicitly state that this was a manual structural verification, and do not claim that the script was run. An oral statement that the profile has been activated is not sufficient.
If the profile cannot be read, is not active, or is structurally invalid, display the following notice and stop personalized topic selection and creation of new tasks. If only the reference location is missing, ask for it first. Do not activate the profile yourself, and do not offer a “generic scheduled task” that borrows private data.

The personal DNA has not yet been activated. Currently, only the general method can be invoked; personalized topic selection, real-experience mapping, and personal-expression calibration remain unavailable. Complete the “Personal Background Interview” and confirm the profile before the full Yetong content-method engine can be unlocked.

The coach's role, service area, local audience, genuine concerns, related business, and publicly shareable experiences and service actions are the basis. If the profile is active but lacks current information about the key audience or city, ask only for that gap; an older but valid
```
schema_version: 2
```
does not require a full re-interview. Do not infer pain points from labels such as “woman” or “mom.” Read the item-by-item permission rules in
```
yetong-topic/references/profile-contract.md
```
first; if they cannot be obtained, follow the permission boundaries in this section and the hotspot rules.

Topic-Selection Rules and Dependencies

Each run must reread the currently bound profile and the Hotspot Retrieval, Screening, and Output Rules; stale retrieval results must not be reused.

Prioritize reading the installed
```
yetong-topic/SKILL.md
```
and
```
yetong-topic/references/topic-method.md
```
, and reuse the Eight Elements and four content-method types defined there. This module's fixed topic count, trend-evidence requirements, and output format take precedence over the regular topic-selection defaults of eight topics, a full candidate table, or a consultation-oriented approach.
If the current task has verifiable Eight Elements Skill or project rules bound to it, first check whether they implement this method and are at the current version, and then read them. A similar name is not sufficient to claim that the resources were read.
If the actual method resources cannot be obtained, use the embedded Eight Elements and four content-method rules, write
```
Rule status: executing under embedded rules
```
, and explain that the method resources were unavailable. Do not claim that a missing Skill was invoked, and do not automatically install dependencies.
The fixed goal is to build professional trust and increase relevant exposure among the coach's local prospective members. Cross-industry trends and non-local reach are allowed, but every topic must answer “Which type of my target member, in what situation, can understand or solve what specific problem?” and explain “Why is this suitable for me, this coach, to discuss?” It must not be framed as teaching other fitness coaches how to market; every topic does not need to sell the course, include a purchase call to action, or be optimized for conversion.

Execution

Trial run / run: Profile passes → perform a live web search and verify timing and data → form six to nine candidates within the scope actually retrieved → screen out three distinct questions, prioritizing relevance to the audience and business → deliver only the final cards. If the information cannot be verified or there are too few candidates, fall back according to the reference; do not fabricate candidates, rankings, or metrics.

Create: First read Native Scheduling and Private-Data Binding. Register a task only when explicit authorization for this creation, complete schedule configuration, an available native scheduling capability, and a future execution environment that can actually access the personal profile and methods are all in place. Prefer a trial run first to check whether the content and references work; a trial run cannot substitute for successful scheduling. If the host has no native capability, provide a copyable task instruction and the remaining conditions, explicitly state “Not created,” and do not write a system cron job, set up a background daemon, or borrow another platform's tool.

Output and Stopping Conditions

One-off content output must contain only the final three topics and the necessary supporting evidence under the hotspot rules. Do not show the full candidate pool or expand the output into body copy, openings, scripts, storyboards, or covers; do not publish content or contact clients.
Confirm successful creation only after native scheduling has succeeded and a verifiable task identifier, enabled state, time and frequency, and time zone have been obtained. Then return the task name, identifier, enabled state, execution time zone, recurrence frequency, next run time, and verification evidence. If any of this creation evidence is missing, write “Creation result unconfirmed” and do not create the task again. If the next run time cannot be obtained, write “Not available”; do not present an estimate as a platform-returned value, and do not claim that execution at the scheduled time has been accepted.
If creation fails, explain the reason, which preparations have been completed, and which configuration is missing. A task card or prompt is not the same as a created task. If the result is unknown, first check the native task list and do not create a duplicate.
Every run must recheck the active status and each item's publication permissions. If the data is invalid, stop that personalized output and prompt the user to repair it; do not alter the profile or pause or delete the native task. Private DNA, task configurations, historical topic selections, member data, and test records must not be placed in a public Skill directory or GitHub.