The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes bash scripts in the 'Preamble' and 'Telemetry' sections of SKILL.md. These scripts are intended to be executed by the agent to perform system enumeration (using uname), directory creation, and configuration file parsing (using cat, grep, and sed).\n- [DATA_EXFILTRATION]: Automated telemetry logic transmits system metadata, including OS and architecture information, to an external endpoint retrieved from the user's local configuration file. This egress occurs via curl POST requests when telemetry is enabled.\n- [EXTERNAL_DOWNLOADS]: The skill's documentation references and encourages the installation of external dependencies, specifically the '@defillama/api' Node.js package and the 'defillama-sdk' Python package. It also relies on network communication with several DefiLlama API domains (api.llama.fi, yields.llama.fi, stablecoins.llama.fi).\n- [PROMPT_INJECTION]: The skill fetches data from external DefiLlama API endpoints. The ingestion of this untrusted data into the agent's context—which also possesses shell execution capabilities—represents an indirect prompt injection surface, as there are no explicit sanitization steps or boundary markers defined for the processed content.

defillama-research