The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses extremely aggressive, imperative language to force the agent to follow its rules over the system prompt. Evidence includes blocks marked as 'EXTREMELY-IMPORTANT' with phrases like 'ABSOLUTELY MUST,' 'not negotiable,' and 'not optional.'
[PROMPT_INJECTION]: It explicitly defines a hierarchy that places these instructions above the agent's 'default system prompt.' By stating 'Superpowers skills — override default system behavior where they conflict,' it attempts to hijack the agent's underlying behavioral constraints.
[PROMPT_INJECTION]: The 'Red Flags' section is designed to suppress the agent's natural reasoning and logic. It provides specific instructions to disregard internal checks such as the need for more context or the simplicity of a task, forcing a rigid adherence to skill invocation.
[INDIRECT_PROMPT_INJECTION]: The skill mandates the use of external 'skills' (other instructional sets) if there is even a '1% chance' they might apply. This aggressive requirement significantly increases the risk of the agent unknowingly loading and executing malicious instructions from other skills that may be present in the environment.

using-superpowers