The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The README.md and install.sh files promote the 'curl | sh' pattern to download and immediately execute code from the 'BofAI/skills' GitHub repository. This method lacks integrity verification and is a high-risk practice for remote code execution.
[PRIVILEGE_ESCALATION]: The install.sh script performs a global installation of the '@tron-walletcli/wallet-cli' NPM package. This action typically requires administrative privileges and affects the global system environment, which is excessive for a localized agent skill.
[EXTERNAL_DOWNLOADS]: The skill requires downloading software from unverified external sources, including a specific NPM package and a GitHub repository, neither of which are on the trusted vendor list.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the TRON blockchain, creating a vulnerability surface. 1. Ingestion points: The skill fetches account history, transaction status, and contract metadata (references/commands.md). 2. Boundary markers: The skill instructs the agent to use '-o json' but lacks explicit instructions to ignore embedded commands within blockchain data fields like memos. 3. Capability inventory: The skill can broadcast transactions and perform global package management. 4. Sanitization: No explicit instructions for sanitizing blockchain-derived content.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its primary functionality. While safety rules are provided to handle passwords via stdin, the underlying mechanism of invoking CLI tools with complex arguments is inherently risky.

wallet-cli