Vai al contenuto
TakoAPI
Torna alle skill

crypto-self-learning

di totaleasy

Coding Agents & IDEs
4.3kDownload riportati da clawskills.sh
11Stelle
1Visualizzazioni
0 likeAccedi per mettere like a questa skill
—Non ancora valutato
La tua valutazione

Sintesi

Self-learning system for crypto trading.

Descrizione

Cosa fa questa skill

Logs crypto trades with full context including technical indicators and macro conditions, analyzes win/loss patterns across dimensions like day of week, RSI ranges, and leverage, and extracts actionable trading rules. Rules are written back to an agent memory file, creating a feedback loop that refines decisions over time.

Rather than manually reviewing trade notes in a spreadsheet, it extracts statistically grounded rules from structured trade history and writes them directly into agent memory for use in future decisions.

Quando usarla

  • Logging a BTCUSDT long trade with RSI and MACD values right after closing
  • Identifying which days of the week have the highest win rate across 50+ trades
  • Generating AVOID and PREFER rules once enough trade history accumulates
  • Running a weekly review to compare current performance against the prior week
  • Auto-updating agent memory with data-backed rules before the next session

Esempio di workflow

INPUT

User asks: I just closed a BTCUSDT long at +1.92% with RSI 28 and a bullish MACD cross. Log it and refresh my rules.

AGENT
  1. 1

    Runs log_trade.py with symbol, direction, entry, exit, pnl_percent, indicators JSON, and market context JSON

  2. 2

    Runs analyze.py to compute win rates by direction, day of week, RSI range, and leverage level

  3. 3

    Runs generate_rules.py to extract AVOID, PREFER, and CAUTION rules from detected patterns

  4. 4

    Runs update_memory.py to append a Learned Rules section to MEMORY.md

OUTPUT

MEMORY.md updated with rules such as 'AVOID: LONG when RSI > 70 (win rate: 23%, n=13)' and 'PREFER: SHORT on Mondays (win rate: 78%, n=9)'

Installazione

clawhub install totaleasy/crypto-self-learning

Esegui il comando CLI oppure incolla il prompt in Claude Code, Cursor, Codex o nell'agente di coding che usi.

Link