The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the Weco CLI by downloading scripts directly from the vendor's official domain (weco.ai). These are verified vendor resources.
[COMMAND_EXECUTION]: The skill uses the weco CLI and executes dynamically generated evaluation scripts. These operations are core to the optimization functionality and are managed via a controlled directory structure (.weco/).
[CREDENTIALS_UNSAFE]: The skill adheres to security best practices regarding secrets. It includes explicit instructions to never read, display, or manually handle .env files, relying instead on standard library loading patterns.
[PROMPT_INJECTION]: Instructions include specific safety guards to ensure the agent treats content within optimized source code or external datasets as data rather than instructions, mitigating indirect prompt injection risks.
[REMOTE_CODE_EXECUTION]: Evaluation templates occasionally reference well-known, trusted data sources (e.g., scikit-learn datasets, Hugging Face) for benchmarking purposes, which is documented as a safe and standard practice for machine learning tasks.

weco