The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses extremely forceful language to override default agent behavior. Examples include: 'YOU ABSOLUTELY MUST invoke the skill', 'YOU DO NOT HAVE A CHOICE', 'This is not negotiable. This is not optional. You cannot rationalize your way out of this.' This pattern is designed to bypass the agent's internal decision-making process regarding when to use specific tools.
[INDIRECT_PROMPT_INJECTION]: By mandating that the agent invoke any skill with even a '1% chance' of applicability BEFORE responding to the user or asking clarifying questions, it creates an aggressive execution surface. If a project contains a malicious skill, this 'using-superpowers' skill ensures that malicious skill is loaded and executed as early as possible in the conversation flow.
Ingestion points: SKILL.md (via the prompt to check all available skills).
Boundary markers: Absent. The skill encourages following external skill content 'directly' without mentioning validation or safety checks.
Capability inventory: The skill itself does not call subprocesses, but it acts as a gatekeeper that forces the invocation of the Skill tool, which loads other content into the context.
Sanitization: Absent. There is no instruction to validate the safety of the skills being invoked.

using-superpowers