The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of computational experiments through the get_results() method, which runs code to perform data analysis and generate visualizations based on a methodology.
[EXTERNAL_DOWNLOADS]: Installation instructions refer to the denario package from a registry and provide links to a GitHub repository and a Docker image.
[CREDENTIALS_UNSAFE]: The documentation guides users on setting up and managing API credentials for LLM providers using environment variables, .env files, and JSON service account keys. These instructions follow standard configuration practices for developers.
[PROMPT_INJECTION]: The skill ingests untrusted content through several input methods, creating a surface for indirect prompt injection.
Ingestion points: Data descriptions, research ideas, and methodology files are ingested via set_data_description(), set_idea(), and set_method() as documented in SKILL.md and references/research_pipeline.md.
Boundary markers: No specific boundary markers or instructions to ignore embedded commands are documented for these inputs.
Capability inventory: The skill can execute arbitrary code in the get_results() stage, write files to the project directory, and interact with network APIs for literature searches.
Sanitization: No explicit sanitization or validation of the ingested research content is described before it is processed by the agents or executed.

denario