The Agent Skills Directory

[SAFE]: The skill implements significant safety controls and a hard boundary policy prohibiting the use of personal health information (PHI) or real clinical decision-making. All processing is deterministic and local, with no network access, external API calls, or LLM-based generation of medical advice.
[COMMAND_EXECUTION]: The skill includes several Python scripts for generating research artifacts such as cohort tables and evidence profiles. These scripts are intended to be run by the user in a local environment using a shell. They are dependency-free, use the standard library, and include strict path validation to prevent access to non-local files or sensitive system areas.
[SAFE]: The documentation includes a 'Security Validation Record' that notes the removal of all previously identified risks, such as network exfiltration and credential handling. The current implementation relies on static structural checks and aggregate data binning as defense-in-depth measures.

clinical-decision-support