The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references the installation of the longbridge-terminal CLI and several standard Python libraries for data analysis. Evidence: Mentions pip install scikit-learn, pandas, numpy, statsmodels, scipy, and arch, as well as the longbridge-terminal tool.
[DYNAMIC_EXECUTION]: The skill documents the use of the longbridge quant run command, which transmits and executes user-defined scripts (Navi or PineScript) on a remote server for quantitative analysis. Evidence: references/quant-cli.md describes the run subcommand for server-side script execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes external financial market data, which introduces an attack surface for indirect prompt injection, although the risk is mitigated by the analytical nature of the tool and instructions to verify data sources. 1. Ingestion points: Market data (K-lines, trades, constituents) fetched via longbridge kline, longbridge trades, and longbridge constituent. 2. Boundary markers: The skill advises citing limitations and verifying data timestamps and coverage. 3. Capability inventory: The skill performs data analysis, script execution via CLI, and local Python-based modeling. 4. Sanitization: The skill instructs the agent to redact credentials and private account data from reports.

longbridge-quant