The Agent Skills Directory

[SAFE]: The skill implements a read-only summarization feature with explicit instructions to avoid modifying, clearing, or compacting the conversation history. This prevents the agent from being used to bypass history-based constraints.
[SAFE]: Platform-level security configurations (disable-model-invocation: true and allow_implicit_invocation: false) effectively restrict the agent's capabilities, preventing it from executing other tools or running without user intent.
[SAFE]: The provided test suite validates safety contracts, ensuring that the skill's instructions maintain strict boundaries on output format and behavior.

recap