The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to interactively develop and execute shell scripts and Python strategies using the Bash tool. The instructions in SKILL.md explicitly direct the agent to run code in a conversational loop, which involves continuous command execution within the environment.
[REMOTE_CODE_EXECUTION]: The documentation in references/indicators.md provides instructions for downloading and building the TA-Lib C library from SourceForge (prdownloads.sourceforge.net) using wget followed by compilation and installation commands (./configure, make, sudo make install). While TA-Lib is a well-known industry library, this pattern involves manual download and execution of source code from an external repository.
[EXTERNAL_DOWNLOADS]: The skill performs legitimate network operations to official services: scripts/fyers_symbols.py fetches daily market symbol master files from FYERS' public infrastructure (public.fyers.in), and scripts/fyers_login.py interacts with the official FYERS API (api-t1.fyers.in) for OAuth authentication.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted market data (historical candles, quotes, and option chains) from external FYERS API endpoints to process signals and display results to the user. 1. Ingestion points: Market data endpoints in scripts/fyers_client.py (history, quotes, depth, option_chain). 2. Boundary markers: Absent; external data is processed and rendered directly in chat. 3. Capability inventory: The skill has access to Bash, Write, and WebFetch tools, creating a significant action surface for potential data-driven attacks. 4. Sanitization: There is no explicit sanitization for the external market data before it is presented to the user.

fyers-trading