The Agent Skills Directory

[OBFUSCATION]: The installation instructions for the Cursor IDE in references/mcp-install.md contain a Base64-encoded configuration string within a deep link. This decodes to a standard JSON object containing the required MCP server URL.
[INDIRECT_PROMPT_INJECTION]: As a research-focused skill, it possesses an inherent surface for indirect prompt injection. It ingests data from external sources (TradingView MCP payloads and Web Search results such as SEC filings and transcripts) and utilizes powerful capabilities like Python data processing, and DOCX/XLSX generation.
Ingestion points: External financial data via tradingview_* tools and narrative content from Web Search (files: SKILL.md, references/workflows/*.md).
Boundary markers: The instructions recommend citing sources and using structured data, but do not specify technical delimiters to encapsulate untrusted text.
Capability inventory: Full use of Python (matplotlib, pandas, seaborn), and the platform's DOCX and XLSX skills for generating institutional reports.
Sanitization: No explicit sanitization or instruction-escaping mechanisms are described in the provided files.
[COMMAND_EXECUTION]: The skill provides installation commands for adding the TradingView MCP server to various IDEs (Cursor, VS Code, Claude Code, etc.) in references/mcp-install.md. These are standard configuration commands intended for user execution to set up the environment.

equity-research-analyst