The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs highly imperative and coercive language (e.g., "EXTREMELY-IMPORTANT", "ABSOLUTELY MUST", "not negotiable", "not optional") to force the agent to follow a specific tool-invocation workflow over its own reasoning.
[PROMPT_INJECTION]: It explicitly instructs the agent to ignore its own judgment or logic, labeling such reasoning as "rationalizing" and providing a list of "Red Flags" to suppress natural instructional responses.
[PROMPT_INJECTION]: The "1% chance" rule and the requirement to check for skills before asking clarifying questions are designed to override the agent's default safety and operational guidelines, potentially increasing the attack surface by forcing the execution of other skills.

using-superpowers