The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute a shell script from a remote URL using a dangerous piped command: curl -sSL https://raw.githubusercontent.com/6551Team/openskills/main/skills/opentrade/install.sh | sh. This pattern is invoked during the initial setup, every 12 hours for updates, and as a recovery mechanism if a command fails.
[EXTERNAL_DOWNLOADS]: Fetches executable content from a third-party GitHub repository (6551Team/openskills). Although the repository appears to be owned by the skill's author, executing remote scripts without verification is a high-risk behavior.
[COMMAND_EXECUTION]: The skill performs several local system operations, including checking for executable paths (which opentrade), reading local state files (~/.opentrade/last_check), and executing the opentrade CLI with arguments derived from user input (e.g., wallet addresses).
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. It ingests data from external sources (blockchain token metadata and wallet balances) and interpolates this data into the agent's context. There are no defined boundary markers or sanitization steps to prevent malicious instructions embedded in token metadata from influencing agent behavior.

opentrade-portfolio