- CONTRIBUTING.md with dev setup, code standards, and difficulty-tiered roadmap
- GitHub issue templates (bug report + feature request) with YAML forms
- README: add Contributing section with contribution areas table
Restructured ml-strategy SKILL.md from a skills-creator perspective:
- Added validate_data() input gate (columns, min rows, NaN ratio)
- Guarded RSI/Bollinger Band division-by-zero with .replace(0, nan)
- Added inf→NaN sanitization at end of build_features()
- Added sliding window option (window_type + sliding_size params)
- Added output contract: fillna(0).clip(-1, 1) guarantees clean signals
- Provided complete SignalEngine example so agent can copy-paste safely
- Reduced Common Pitfalls from 7 to 3 (4 now enforced by code)
- Zero changes to backtest engine — all fixes are skill-only
- Add DataLoader Protocol + Registry with market-level fallback chains
- New data sources: AKShare (A-shares/US/HK/futures, free) and CCXT (100+ crypto exchanges)
- Auto-fallback: tushare unavailable -> akshare; OKX unavailable -> CCXT
- Zero-config: all markets work without any API keys
- New web_search tool (DuckDuckGo) for web information retrieval
- Skill categorization: 68 skills organized into 7 categories (data-source, strategy, analysis, asset-class, crypto, flow, tool)
- New data-routing meta-skill for intelligent data source selection
- Registry self-initializes via _ensure_registered() — no manual imports needed
- Pattern detection returns market type instead of source name (decoupled)
- runner.py loads .env via dotenv for reliable token access
Add full-stack Pine Script export: backtest strategies can now be
converted to TradingView Pine Script v6 via natural language, CLI,
API, or the web UI.
- New `pine-script` skill with Python-to-Pine mapping table and v6 syntax guide
- CLI: `/pine <run_id>` slash command and `--pine` flag
- API: `GET /runs/{run_id}/pine` endpoint
- Frontend: Pine Script button on RunCompleteCard + modal code viewer with copy
- README: updated news, capabilities, skill count (65), CLI/API docs