The Agent Skills Directory

[SAFE]: The skill is composed entirely of natural language instructions intended to guide the agent in drafting interpersonal communications. It does not contain any executable code, scripts, or system-level commands.
[NO_CODE]: The skill does not include any Python, Node.js, or shell scripts. It relies solely on the underlying agent's text generation capabilities.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided context regarding recipients and relationships. Although this represents an input surface for potential indirect injection, the skill includes explicit procedures for the agent to restate boundaries and confirm assumptions with the user before finalizing drafts, which serves as a safety mechanism. Furthermore, the skill has no capabilities to perform network operations or file modifications based on that input.

charm-person