The Agent Skills Directory

[SAFE]: The skill consists entirely of markdown instructions and templates for text generation. It does not include any scripts, dependencies, or network-enabled tools.\n- [INDIRECT_PROMPT_INJECTION]: The skill accepts clinical context and instructions as input. While it lacks specific boundary markers to prevent instructions embedded in the clinical data from being followed, the skill lacks access to any dangerous tools, network connections, or sensitive files that could be exploited. The risk is limited to the generated text draft, which the skill explicitly mandates must be reviewed by a human professional.\n- [METADATA_POISONING]: The metadata correctly describes the purpose of the skill, and no deceptive instructions are present in the headers or descriptions.

patient-communication