The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests external market data from multiple APIs, creating a surface for indirect prompt injection through data poisoning.
Ingestion points: External OHLCV data is fetched via client.history() (OpenAlgo API), yf.download() (Yahoo Finance), and exchange.fetch_ohlcv() (CCXT/Binance) in strategy scripts such as rules/assets/ema_crossover/backtest.py and rules/assets/supertrend/backtest.py.
Boundary markers: Absent. The skill does not employ delimiters or specific instructions to the agent to ignore potentially malicious instructions embedded in the external market data.
Capability inventory: The skill has network access for fetching data, file writing capabilities for exporting trade logs to CSV, and visualization capabilities via Plotly.
Sanitization: Absent. Market data and derived metrics are interpolated directly into generated reports and explanations without specific sanitization or validation filters.
[EXTERNAL_DOWNLOADS]: The skill documentation recommends the installation of additional third-party Python packages to extend its capabilities.
Evidence: Instructions in rules/quantstats-tearsheet.md recommend pip install quantstats. Documentation in rules/data-fetching.md suggests installing ccxt, alpaca-trade-api, and twelvedata.
Context: These packages are well-known and widely used in the quantitative finance community.

vectorbt-expert