FlowCay logo
FlowCay
LytSkill Cross-border Analytics

LytSkill Cross-border Analytics

Skill focused on cross-border e-commerce business analysis and operational insights.

Install command

npx skills add https://github.com/YYYYYZhao/Lytskill
SkillOpen SourceAll Platforms
Added on Popularity 1

Details

Overview

Lytskill is a suite of Agent skills for TikTok Shop and cross-border e-commerce operations (current version v2.0.0). It works inside AI coding/agent tools that support skill or system-prompt capabilities, including Claude Code, Codex, Cursor and Trae Solo. Rather than giving "one-size-fits-all answers," it breaks vague business problems into diagnosable questions, turns product-selection judgments into testable hypotheses, converts product facts into shootable short-video scripts, and transforms backend data into concrete next actions. It is a free open-source project hosted on GitHub at YYYYZhao/Lytskill.

Key Features

  • 13 skills in one repo (v2.0.0) covering the full cross-border e-commerce workflow:
  • lyt — main router: judges the business boundary and routes to the correct specialized skill when the user is unsure.
  • lyt-problem-clarifier — when a user asks "why no orders / why no traffic / why ads don't spend", asks structured follow-up questions and identifies the data breakpoint.
  • lyt-traffic-logic — explains the underlying distribution logic of store, product, product card, short-video and ad traffic.
  • lyt-product-selection — from-zero product selection: category/country/traffic-channel suitability and building a candidate pool.
  • lyt-product-validation — validates existing candidate products (screenshots, supply-chain docs, test data) to decide whether they enter the testing pool.
  • lyt-page-diagnosis — diagnoses main images, secondary images, detail pages and first screens; localizes product-message, click-reason, trust-evidence and buyer-doubt issues.
  • lyt-detail-page (new in v2.0.0) — generates a complete product detail page from zero using product photos, packaging, spec sheets and visual references; outputs page structure, per-screen copy, visual notes, image prompts, independent splits and a full long image.
  • lyt-image-patterns — batch-analyzes competitor/Top-store main images to extract testable main-image expression formulas.
  • lyt-image-prompt — reverse-engineers an e-commerce reference image and merges the user's own product into its composition and style.
  • lyt-video-script — generates shootable storyboards, lines, on-screen text, shot lists and content-validation hypotheses for product short videos.
  • lyt-data-analysis — decomposes GMV variables from screenshots, CSV/XLSX and backend data, then outputs diagnostics and next actions.
  • lyt-learning — breaks a business topic into a continuous learning series that adapts based on feedback.
  • lyt-knowledge-check — verifies understanding with real scenario questions covering conclusions, evidence, causal chains, applicability and minimal validation actions.
  • Detail-page generation workflow: plan version, production version and final version; brand/product/activity slogans (keeps user copy verbatim, offers ≤3 candidates only when generation is needed); first-round info collection uses a filled-in example template; the final version confirms each independent split before stitching the full page.
  • Router + handoff: all skills follow unified naming (≤3 words) with updated handoff paths across selection, validation, diagnosis and analysis skills.

Install & Update

  • Full install: npx -y skills add YYYYZhao/Lytskill -g --all
  • Claude Code plugin: claude plugin marketplace add YYYYZhao/Lytskill then claude plugin install lyt@lytskill; update via claude plugin marketplace update lytskill + claude plugin update lyt@lytskill
  • Trae Solo: download the packaged zip from Releases and import each skill pack individually.
  • Local build: bash tools/build-skills.sh outputs to dist/skills/ (build script auto-discovers all skills under skills/, no fixed manifest to maintain).

Best For

TikTok Shop and cross-border e-commerce operators who want their AI agents to follow real business logic — diagnosing why a product isn't selling, understanding traffic allocation, selecting/validating products, generating detail pages and short-video scripts, and turning backend data into verified next actions.

FAQ

What is Lytskill?

A free, open-source set of 13 Agent skills for TikTok Shop / cross-border e-commerce operations that can be used inside Claude Code, Codex, Cursor, Trae Solo and similar skill-capable agents.

What does v2.0.0 add?

A new lyt-detail-page skill (from-zero product detail page generation with independent splits and full long image) and shortened names for 4 existing skills (e.g. lyt-image-to-prompt → lyt-image-prompt); the older names are no longer kept.

How do I install it?

npx -y skills add YYYYZhao/Lytskill -g --all, or add the repo as a Claude Code plugin marketplace and install lyt@lytskill.

Does it guarantee sales or profits?

No — it explicitly does not promise revenue or "viral/big-money" outcomes; platform rules, commissions, logistics and restricted-sale policies must be verified against official/latest sources, and conclusions stay conservative when data is limited.