The Agent Skills Directory

[SAFE]: The skill is entirely instructional and focuses on altering the agent's linguistic style. It does not perform any file system operations, network requests, or credential handling.
[SAFE]: The skill includes instructions to maintain a persistent persona across conversation turns ('ACTIVE EVERY RESPONSE. No revert after many turns'). While this is a form of instructional persistence, it is used here solely for stylistic purposes and does not attempt to bypass safety filters or override system-level security constraints.
[SAFE]: The configuration in SKILL.md uses disable-model-invocation: true, which is a security-positive setting that prevents the agent from calling tools or executing functions while the skill is active, thereby limiting the potential attack surface.
[SAFE]: The skill provides explicit safety boundaries, instructing the agent to drop the caveman style when presenting security warnings or irreversible action confirmations to ensure clarity and safety.

caveman