The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses high-pressure, imperative language designed to override the agent's internal decision-making processes. Phrases like 'NON-NEGOTIABLE RULE', 'YOU DO NOT HAVE A CHOICE', 'You cannot rationalize your way out of this', and 'STOP—you're rationalizing' are used to force a specific behavior (mandatory skill checking) regardless of the agent's assessment of the task's complexity.
[PROMPT_INJECTION]: The instructions explicitly lower the agent's heuristic threshold for tool usage, mandating invocation if there is even a '1% chance' a skill applies. This is a directive to bypass the agent's standard logic for determining when a tool is necessary or appropriate.
[INDIRECT_PROMPT_INJECTION]: By acting as a meta-skill that governs 'any conversation,' it establishes a persistent instructional framework that influences the agent's processing of all subsequent untrusted user inputs. The mandatory invocation workflow increases the attack surface by forcing the agent to actively search for and potentially execute other skills based on minimal or adversarial triggers in user messages.

using-superpowers