The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses extreme imperative language and non-negotiable commands, such as 'ABSOLUTELY MUST', 'YOU DO NOT HAVE A CHOICE', and 'This is not negotiable', to force the agent into a specific behavior pattern.
[PROMPT_INJECTION]: It explicitly instructs the agent to bypass its own internal reasoning and safety filters, stating 'You cannot rationalize your way out of this'.
[PROMPT_INJECTION]: The 'Red Flags' section provides a list of common logical and safety-oriented thoughts (e.g., 'I need more context first', 'This doesn't count as a task') and labels them as rationalizations to be ignored.
[PROMPT_INJECTION]: By requiring skill invocation even when there is only a '1% chance' of applicability, the skill creates a massive vulnerability surface where the agent is forced to ingest and execute potentially untrusted external instructions before any safety context is established.
[PROMPT_INJECTION]: Ingestion points: The skill instructs the agent to use the 'Skill' tool to load external files. Boundary markers: None identified; the skill mandates immediate obedience. Capability inventory: The skill mandates use of the 'Skill' tool, which has the capability to load and execute other skill instructions. Sanitization: No sanitization or validation of the invoked skills is mentioned or required.

using-superpowers