The Agent Skills Directory

[COMMAND_EXECUTION]: The documentation specifies several shell commands for system setup and operation, including package installation using uv and pip, system-level package installation via sudo apt-get install texlive-full, and launching the application's graphical interface with denario run.
[EXTERNAL_DOWNLOADS]: The skill references external resources for installation, including the official source repository on GitHub (AstroPilot-AI/Denario) and a pre-configured Docker image (pablovd/denario) on Docker Hub.
[REMOTE_CODE_EXECUTION]: A core component of the skill is the 'Execution Agent,' which dynamically generates and runs Python code. As described in references/research_pipeline.md, the den.get_results() method performs the actual computational work by executing code to conduct analyses, create visualizations, and compile findings. This dynamic code execution occurs at runtime within the user's environment.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. It ingests untrusted data through the den.set_data_description() method (as seen in references/research_pipeline.md), which is then used by the 'Idea Agent' and 'Method Agent' to determine research goals and execution logic. There are no explicit boundary markers or sanitization processes mentioned to prevent malicious instructions within these descriptions from influencing the generated code execution in later stages.

denario