The Agent Skills Directory

[SAFE]: The skill is a set of instructional markdown files defining a mathematical framework for action selection. It does not contain any executable code, shell scripts, or binaries.
[EXTERNAL_DOWNLOADS]: No external downloads or remote dependencies are defined in the skill configuration or instructions.
[DATA_EXFILTRATION]: While the skill logic involves processing task state and observations, it does not include any tools or instructions for performing network requests or accessing sensitive local files.
[PROMPT_INJECTION]: The instructions provide a specific orchestration logic ('prescriptive mode') but do not attempt to override system safety guidelines, bypass ethical filters, or extract system prompts.
[COMMAND_EXECUTION]: The skill mentions actions like tool_call and retrieve, but these are abstract placeholders for the agent's existing capabilities; the skill does not execute arbitrary shell commands or provide unsafe execution environments.

utility