x-trader-skill-builder

name	x-trader-skill-builder
description	Build trader-specific research-model agent skills from public X/Twitter post datasets, exported timelines, saved threads, signal CSVs, and quote-aware review files. Use when an agent needs to reverse-engineer an X trader's investment logic, separate forward-looking thesis posts from quote-only or retrospective performance posts, generate clean signal datasets, derive a high-quality thesis template, or prepare a reusable trader-specific agent skill for platforms such as Claude Code, OpenClaw, Codex-style skill systems, or other local AI agent runtimes.
quantSkills	{"organization":"https://github.com/quantskills","repository":"quantskills/skill-x-trader-builder","repository_url":"https://github.com/quantskills/skill-x-trader-builder","project_type":"skill","collection":"trader-research-models","license":"GPL-3.0","category":"tooling","tags":["trader-skill","research-model","x-twitter","skill-builder","workflow"],"platforms":["claude-code","codex","openclaw","cursor"],"status":"stable","validation_level":"runnable","maintainer_type":"official","summary_zh":"把任意 X/Twitter 公开交易员的发帖历史，加工成 trader 专属的研究模型 Skill：init-run → 采集 → extract → auto-review → split → evaluate → template → report 九步流水线，从噪...","summary_en":"Skill-builder workflow for turning public X/Twitter data and user materials into trader-specific research-model skills."}

{
  "version": 1,
  "task": {
    "placeholder": "请提供交易员名称、X/Twitter 帖文导出或数据路径，并说明时间范围与期望产物",
    "required": true
  },
  "fields": [
    {
      "key": "trader",
      "label": "交易员名称",
      "type": "text",
      "placeholder": "公开账号名或研究模型名称",
      "required": true
    },
    {
      "key": "focus",
      "label": "提炼重点",
      "type": "text",
      "placeholder": "如：前瞻信号、主题机制、催化剂、风险与跟踪指标"
    }
  ],
  "prompt_template": "{{#task}}任务与材料：\n{{task}}\n\n{{/task}}{{#attachments}}用户上传的材料（已放入工作区）：\n{{attachments}}\n\n{{/attachments}}请把交易员 {{trader}} 的公开 X/Twitter 帖文历史加工为可复用研究模型{{#focus}}，重点提炼 {{focus}}{{/focus}}；区分本人观点、引用内容、前瞻论点与事后业绩，完成抽取、语义复核、数据集拆分、适用时的前瞻收益评估、论点模板和专属 Skill 素材，输出中文报告。"
}

X Trader Skill Builder

Use this skill to turn a public trader post history into a reusable research model. It generalizes the Serenity MVP workflow: clean noisy public posts, label semantic intent, isolate forward-looking signals, extract high-quality thesis patterns, and prepare the ingredients for a trader-specific agent skill.

Workflow

1. Initialize a real-data run folder.

   • Use scripts/x_trader_builder.py init-run --trader "<name>" --trader-slug <slug> --out real_runs.
   • This creates a Serenity-grade checklist and standard sources/ and outputs/ folders.

2. Collect public data.

   • Accept CSV, JSON, JSONL, TXT, or Markdown exports.
   • Prefer columns: created_at, text, url, ticker, quoted_text, theme, evidence_types, supply_chain_role, engagement_score.
   • Record data source, retrieval date, and missing coverage.
   • For stable collection and normalization, use collectors/stable_collectors.py and the guide references/data_collection_stable_zh.md.
   • Prefer official API exports, user-owned exports, Apify/managed scraper exports, RSS feeds, and public article archives over DIY X page scraping.
   • If the user has already logged into X in a local browser, a browser-assisted fallback can be used with collectors/browser_scroll_cdp.mjs. Launch Edge/Chrome with a local CDP port, navigate logged-in X search/profile pages, and export visible posts into the same posts.csv contract. Treat this as a partial capture unless it is date-windowed and reviewed.

3. Extract raw public signals when only post exports are available.

   • Use scripts/x_trader_builder.py extract --posts <posts.csv|json|jsonl|txt|md> --trader "<name>" --trader-slug <slug> --out <run>/outputs/raw_extract.
   • This creates signals.csv, no_ticker_theme_posts.csv, and extract_summary.md.

4. Run semantic review.

   • Use scripts/x_trader_builder.py auto-review --signals <signals.csv> --out <dir>.
   • This labels rows as keep, keep_deweighted, deweight, delete_from_this_signal, remove_from_forward_signal_keep_as_track_record_context, keep_as_explainer_deweight, or delete.

5. Split the dataset.

Interpretation Rules

• Treat public posts as research artifacts, not audited P&L.
• Separate a trader's own words from quoted text.
• Do not treat retrospective return claims as forward signals.
• Keep watchlists and crowdsourced lists, but deweight them.
• Keep broad baskets, but deweight single-ticker signal strength.
• Give highest weight to posts with mechanism, evidence, risk, and tracking logic.

Git Hygiene

Generated CSVs, source checkouts, and price files are not skill dependencies. Keep large run artifacts out of the skill folder and usually out of Git. Upload concise Markdown reports, schemas, and scripts; store large data separately if needed.