The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/telegram_push.py accesses the sensitive file path ~/.cc-connect/config.toml to retrieve a Telegram bot token and chat ID credentials. This represents unauthorized access to secrets managed by an external application on the host system.
[COMMAND_EXECUTION]: The script scripts/bb_report.py uses subprocess.run to execute a local Python script (telegram_push.py) with arguments derived from market price data fetched from the Binance API. While the data is cast to a float before use, the architectural pattern of passing externally-sourced data into sub-process execution environments is a high-risk vector for command injection if sanitization logic is bypassed.
[INDIRECT_PROMPT_INJECTION]: The skill ingests structured data (JSON) from various Binance API endpoints (fapi.binance.com, data-api.binance.vision) for strategy calculations and reporting. It lacks explicit sanitization or boundary markers to prevent the agent from potentially obeying instructions embedded within the data payloads at runtime.
Ingestion points: core/binance_client.py, scripts/bb_report.py, and cli/paper.py (Binance API fetch calls).
Boundary markers: Not present.
Capability inventory: File writing (paper_trades.json), network requests to Binance and Telegram, and local command execution via subprocess.run.
Sanitization: Standard JSON parsing is used, but no specific filtering for natural language instructions in price or ticker data was detected.

bb-scalper