alchaincyf/x-mentor-skill
Developer toolsThis page was translated automatically. Technical details may differ from the original. View the original
X Tushi.skill — Nuwa's first "non-human" work. Distilling methodologies from 6 top X creators + open-source algorithm data, distilling a complete topic-writing-growth manual. Made with Nuwa.skill
Overview
Based on the methodology of six million-dollar annual X creators—Nicolas Cole, Dickie Bush, Sahil Bloom, Justin Welsh, Dan Koe, and Alex Hormozi—+ precise weighting data from the X open-source algorithm, it distills 6 core mind models, 10 decision heuristics, and a complete topic-writing-growth manual. Why do this: The longer the Skill.md, the easier it is for AI to "get lost in knowledge and forget to act" during actual execution. 249 lines of routing + execution rules let AI get started first, loading as needed when deep knowledge is needed. Process: - Automatically collect data from nearly 100 user tweets via computer-use / browser tool - If automatic collection fails, guide users to manually provide data (three-level downgrade) - Generate economist-style HTML diagnostic reports (KPI/content ROI/communication funnel/time analysis/brand narrative/action recommendations) - All data and reports are saved in user-data/{username}/ directory Skill can now "remember users." Each time activated: - Automatically checks whether the user's historical diagnostic data is present - If → silently reads personalized strategies to make recommendations more accurate across all scenarios - such as re-diagnosis after more than 30 days → reminders We conducted real-world A/B testing on v2—the same prompt was run with two sub-agents: one with skill and one without: The quality of the tweets themselves isn't much different (good LLMs can write decent tweets), but **—know why it's written this way, when to publish, and what happens to the data after posting. This skill is based on the open Agent Skills protocol and can run on any skills-compatible AI agent runtime (Claude Code, Codex, Cursor, OpenClaw, Hermes Agent, CodeBuddy, Workbuddy, Gemini CLI, OpenCode, etc., 50+ runtimes). The universal CLI installer (vercel-labs/skills, supporting 55+ runtimes) will automatically recognize the current runtime and place skill in the correct directory. When specifying runtime, you need to add parameters such as -a claude-code / -a codex / -a cursor / -a openclaw when specifying runtime.
Overview (machine translated)
The animation is produced by huashu-design skill
Use your own account data to correct your writing style, making real experiences and reusable value the core of your content.
**Nuwa’s first ‘non-human’ work. It’s not about distilling a person, but about distilling a field. **
Based on Nicolas Cole, Dickie Bush, Sahil Bloom, Justin Welsh, Dan Koe, Alex Hormozi
Methodologies of six creators + Tibo’s hands-on review + first-hand information on the X platform,
Distills 6 core mental models, 10 decision heuristics, and a complete topic-writing-growth manual.
Usage Example · Installation · This update · What was distilled · Research source
Updated on 2026-10-11
Incorporate Tibo’s content review into the core guide: using your own data to correct writing habits, valuing complete expression, real cost, reusable guides, personal experience, and sustained investment. Author cases and platform facts are marked separately, without directly copying English market results.
- Removal of hard restrictions such as 120–130 characters, mandatory reply for external links, default 60% English, and fixed thread length.
- No longer interpreting the weight of algorithm prediction probabilities as “one reply equals how many likes.”
- 30-minute low performance does not automatically delete posts or repost; By default, observe for 24 hours or 72 hours, then test another variable.
- Account diagnosis distinguishes between main posts, replies, native articles, and external links, distinguishing exposure, attribution of followers, net follower growth, and usage behaviors.
- Increase experience feedback and error correction; Uncle Hua’s expression examples are optional personal style references and are not all users’ personas.
This time, we conducted independent rule trials and statistical reviews, fixing conflicts in reference files. They verify rule consistency and calculation criteria, but do not mean that user growth has been verified through operational experiments. Original account backend data, private policy files, and local reports are not included in the repository; Retain public migration examples and expression examples.
Usage Examples
“I want to write an 800-word review of a Chinese project, with the main text providing the open-source address.”
→ First, verify the actual results and experiences, describe the content completely, and place the link in a convenient position without forcing it into a short post.
“Posting for 30 minutes and not going viral, should I delete and repost?”
→ Don’t treat 30 minutes as a life-or-death line; Record 24/72 hours of performance, check initial commitments and main text delivery, and adjust for new value.
“Analyze my X account and tell me how to post next month.”
→ Using authorized backend or exporting data, reporting samples, time windows, statistical scopes, and missing items, then providing specific actions and verification methods, does not guarantee growth.
Installation
This skill is based on the open Agent Skills protocol and can run on any skills-compatible AI agent runtime (Claude Code, Codex, Cursor, OpenClaw, Hermes Agent, CodeBuddy, Workbuddy, Gemini CLI, OpenCode, etc., 50+ runtimes).
Method 1: One-line command (recommended, automatic detection across runtime)
This page was translated automatically. Technical details may differ from the original.
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add alchaincyf/x-mentor-skill