The Agent Skills Directory

[SAFE]: The skill consists entirely of natural language instructions for a conversational AI and does not include any executable code, shell commands, or scripts.- [DATA_EXPOSURE]: No sensitive file access, credential harvesting, or network communication patterns were identified. The skill is designed to facilitate a private preparatory conversation.- [COMMAND_EXECUTION]: The skill correctly uses the disable-model-invocation: true flag in its metadata, which prevents the agent from calling external tools or executing commands, further hardening its security posture.- [INDIRECT_PROMPT_INJECTION]: The skill ingests user input (descriptions of situations and draft messages) to provide feedback. While it lacks explicit delimiters for this input, the risk is minimal as the skill has no capability to perform actions outside of text generation.

hard-news