The Agent Skills Directory

[SAFE]: The skill defines a purely instructional methodology for the agent to follow during user interactions. It uses a "design tree" approach to structure rounds of questions and answers. There are no signs of prompt injection, data exfiltration, or malicious commands. The instructions regarding "sub-agents" to find facts from the environment utilize the agent's native capabilities and do not involve untrusted external data sources.
[NO_CODE]: The skill does not contain any executable scripts, dependencies, or external binary files. All functionality is handled through the agent's native reasoning and tool-use capabilities based on the provided natural language instructions. The YAML configuration restricts model invocation and implicit execution, which are security-hardening measures.

batch-grill-me