The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data retrieved via TradingView MCP tools such as tradingview_get_news and tradingview_get_calendar. Maliciously crafted headlines or event markers in these external feeds could attempt to influence the agent's output or reasoning process. This is mitigated by the structured analysis-framework.md, which mandates cross-checking conclusions across multiple independent data types.
Ingestion points: references/equity-research.md, references/event-study.md, references/market-macro-research.md.
Boundary markers: The skill uses explicit research modes and a shared evidence level system (Observed, Derived, Interpreted, Actionable) to delimit data processing.
Capability inventory: The skill utilizes a suite of tradingview_* tools for market data retrieval and search.
Sanitization: The workflow requires recording missing data, assumptions, and counter-evidence to validate external inputs.
[DYNAMIC_EXECUTION]: Multiple reference files, including references/pattern-library.md, references/risk-management.md, and references/technical-analysis.md, contain JavaScript code snippets that define quantitative formulas and pattern recognition criteria. These are provided as logic templates for the agent to use when performing calculations or identifying market patterns. The code is transparently documented and limited to financial mathematics and indicator logic.

tradingview-quantitative