Ye Tong Fitness Coach: Xiaohongshu Topic Ideas
Turn the coach's authentic personality and local customer problems into Xiaohongshu topics that can be filmed or written and may attract nearby potential members. The output consists of topic ideas and the basis for creating them—not finished copy, traffic forecasts, or conversion guarantees.
Inputs and Entry Points
The user must explicitly provide or pass in the following:
- The path to the coach's ; do not scan private directories to find the profile;
- The local in-person member segment the coach wants to reach in this round and the topic objective; do not ask again if this is already clear in the profile;
- Optional preferences such as the number of topics, format, or recent assets. If none are specified, default to 8 candidates without requiring each of the eight elements to appear once.
If the user makes a short request such as “Give me 5 more topics” immediately after completing a specific topic or draft, carry forward the core proposition from the previous round by default and create a series extension of the same theme; begin with one sentence explaining what is being carried forward. Only switch to generating topics broadly from the entire profile when the user explicitly asks to change direction or start over, or when the previous theme cannot be reliably identified from the current conversation. Do not repeatedly ask about a known theme merely because the user only provided a number.
First, run
python3 scripts/check_profile.py <person-dna.md>
from the directory containing this Skill, or use the script's actual absolute path; do not treat the user's current project directory as the Skill directory. Then follow the
Personality Profile Reading Contract to verify the source of facts and the disclosure level. Enable personalized mode only when
and the profile is structurally readable. Access is limited to the user-provided profile and the assets explicitly assigned to the task. Do not copy the profile into this Skill directory, test samples, or distributable files.
Use these six groups of profiling data as needed for the task: venue strengths; the local customer segment the coach serves most, including its goals and obstacles; exact customer quotes and real examples of how they were handled; the coach's identity and courses or supplementary services; highlight experiences and anonymized case studies; and the coach's own summary of the reasons for choosing them. Activated profiles using the older
remain usable. Missing new data does not revoke activation or require the user to answer the six groups again; ask a follow-up question only when the current topic genuinely lacks essential evidence.
If the profile does not exist, is structurally invalid, or does not have
, first display:
Personality DNA has not been activated. Right now, only the generic method can be used; personalized topic generation, real-experience mapping, and calibration to the coach's own voice are unavailable. Complete the “Personality Background Interview” and confirm the profile to unlock Ye Tong's full content-method engine.
Provide a “generic version” only after the user explicitly asks to skip activation. Do not use any personal experiences, city, venue, or customer stories, and do not claim that the output sounds like the coach. Do not change the profile status yourself; direct the user back to the Personality Background Interview.
Core Workflow
- Define the local problem. Generate topics only for local potential members near the coach's location: who may notice, inquire about, or book a service in their everyday lives and during gym visits. Even if the profile records other businesses, do not mix in material that teaches peers how to acquire clients. If the user explicitly asks for advice on acquiring peers as clients, explain that this is outside the scope.
- Extract evidence. Prioritize connecting the stated goals and underlying obstacles of the local customer segment the coach serves most with the coach's genuine venue strengths, specialties, class standards, supplementary services, or highlight cases. Each topic should use only the evidence needed to answer its specific question; do not pack the entire Personality DNA into one topic. Distinguish verbatim customer quotes, coach paraphrases, the coach's own assessment, and Agent inferences. Examples given during an interview cannot be treated as customer facts, and items marked “To be confirmed” should become follow-up questions only. See the Personality Profile Reading Contract for evidence-collection rules.
- Generate candidates. First establish the “local audience + specific problem + provable perspective of the coach,” then choose one primary breakout element and, if necessary, one secondary element. See Topic Method for the boundaries and combination rules of the eight elements. When extending an existing theme, break the same core proposition into different member questions, class moments, or examples of service evidence. Do not merely rewrite the same sentence five times or jump to an unrelated theme without reason. Do not fabricate stories, chase trends, or manufacture extreme conflict merely to include every element.
- Assign a content type. Mark only one primary type for each topic: Persona type, Marketing type, Altruistic type, or Traffic & Interaction type. See Topic Method for the classification rules. If a type lacks genuine material, identify the gap rather than forcing a quota.
- Screen the topics. Review each topic for local relevance, distinctiveness of the coach's persona, authenticity of the evidence, feasibility of filming or writing it, potential consultation value, and disclosure risk. Merely placing a city name before the title does not count as localization. Every topic must connect to a genuine local gym-visit context, a known customer concern, or a service scenario the coach can showcase locally. Prioritize topics that answer real consultation concerns; do not use speculative view counts as a substitute for member-acquisition value.
- Deliver the results. Provide the candidates, the top three topics to prioritize and the reasons for that ranking, the missing evidence, and the topics that require verification before publication. If the user requests a complete post, stop at the topic-selection boundary and hand off to the subsequent copywriting workflow.
Output for Each Topic
By default, use a compact table or numbered cards. Each topic should include:
- A working title and one specific, shootable opening scenario;
- The primary element and optional secondary element, along with its content type;
- The corresponding local audience, specific concern, and profile source;
- The coach's genuine experience, service action, or available evidence. If the evidence is insufficient, mark it as “Evidence needed” and do not present it as ready to publish with certainty;
- Why it may attract nearby potential members and what risks must be checked before publication.
Candidates are not final publication titles. Any result, price, duration, or participant count stated in a title must be traceable to the profile or newly provided evidence. Fields may be condensed when shown to the user, but traceable fact mapping must be preserved.
Invariants and Stopping Conditions
- The user's “one element 50% / two elements 90% / ten million views” statements are rules of thumb, not facts, probabilities, or KPIs. Do not promise viral posts, fat-loss results, or conversions.
- The eight Douyin elements are only a source of creative inspiration. Do not assume that a title effective on Douyin will work on Xiaohongshu. Xiaohongshu topics must return to local search intent, trust-building, and the path to an in-person visit.
- Do not infer income, body-image anxiety, marital or childbearing plans, spending power, or motivation from city, gender, age, or occupation labels.
- Do not disclose member names, photos, chat logs, or health details without authorization. Even when anonymized, check for the risk of reverse identification when small-city and venue information are combined.
- Medical, diet, skin-looseness, weight-rebound, and training-result topics must be framed cautiously. Do not turn a single case study or personal method into a universal guarantee.
- Do not present an unverified service menu, price, refund policy, or metric definition as a confirmed promise. Do not use unverified claims such as “the worst,” “No. 1,” or “double the success rate.”
- If every candidate lacks genuine evidence, provide no more than 2–3 directions explicitly marked “To be validated” and ask the single question most likely to fill the content-evidence gap.