The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill setup process in SKILL.md downloads an installation script from the vendor's official GitHub repository (okx/onchainos-skills). This process includes a security best-practice where the SHA256 checksum of the downloaded script is verified against a checksum file before execution, preventing the execution of tampered code. Automated scanner alerts regarding a piped shell command in scripts/scan_live.py were identified as false positives, as the flagged string is part of a manual instruction printed for the user rather than an executable command.
[COMMAND_EXECUTION]: The bot executes the onchainos CLI tool via Python's subprocess module to perform on-chain operations. These calls are made using list-style arguments without shell=True for trading logic, which mitigates command injection risks. A single instance of shell=True in the dashboard setup is used exclusively with a hardcoded port number and is therefore not exploitable.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted metadata from the Solana blockchain, such as token symbols and descriptions from various launchpads. While this creates a surface for indirect prompt injection, the risk is minimal as the ingested strings are primarily used for display in the local dashboard and numerical analysis by the trading logic rather than being re-interpolated into agent prompts.
[SAFE]: The skill is developed by the vendor (okx) for their own ecosystem (onchainos). All network operations target official vendor domains, trusted platform registries, or local services. The software incorporates extensive human-in-the-loop safety protocols, including mandatory typed confirmation for live trading and autonomous risk gates that pause execution if loss limits are reached.

meme-trench-scanner