The Agent Skills Directory

[SAFE]: The skill acts as a high-level orchestration and planning guide for research experiments. It does not contain executable scripts, hidden commands, or unauthorized network operations. The instructions focus on logical workflows and document generation.
[INDIRECT_PROMPT_INJECTION]: The skill identifies ingestion points for external data, such as reading project logs, dashboards, and past run results to establish baselines. While this processes untrusted data, the skill uses this information only to inform a planning process and does not perform sensitive operations or automated command execution based on that data.
[COMMAND_EXECUTION]: The skill describes a process for generating command lists and configuration prompts for experimental stages. However, it explicitly delegates the execution of these commands to separate, specialized agent sessions and emphasizes that formal execution is subject to researcher permission and project-specific safety rules.

exp-campaign