The Agent Skills Directory

[SAFE]: The skill consists entirely of natural language instructions and contains no executable code, scripts, or binary files. The instructions are focused on re-formatting existing conversational context into a simpler style.
[INDIRECT_PROMPT_INJECTION]: The skill processes the assistant's own previous message as input. While this involves processing data that may have originated from external sources, the skill contains explicit constraints (Rule 1: 'Never answer a new question, never add new information, never use tools') that prevent the agent from escalating privileges or performing new actions. Furthermore, Rule 3 ('Facts survive verbatim') instructs the agent to maintain identifiers and URLs exactly as they were, reducing the risk of the simplification process being used to mask or alter sensitive technical data.

pflow-bro