The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs direct shell interpolation of variables such as $SYMBOL, $PRICE, $D_CLOSES, and $W_CLOSES into python3 -c command arguments. A symbol name or market data point containing shell metacharacters or Python escape sequences could allow an attacker to execute arbitrary commands on the underlying system.
[DYNAMIC_EXECUTION]: The skill uses a hardcoded absolute path to a specific user directory (/Users/engineer/.venv/bin/python3) to execute the indicators.py script. This environment assumption is a security risk as it points to specific host structures and could be exploited to hijack execution flow if the environment is predictable or partially controlled.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted market data and user-provided symbols and processes them through local scripts and shell commands without sanitization.
Ingestion points: symbol lists (user input) and technical data (OHLCV, study values) fetched from the TradingView API.
Boundary markers: None present; the skill lacks delimiters or instructions to ignore embedded malicious content in the fetched data.
Capability inventory: Shell execution (python3), filesystem access (mkdir), and file write operations (open().write()).
Sanitization: None; the skill does not perform escaping, validation, or filtering on the data before using it in executable contexts.

tradingview-fetch