The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs authoritative and absolute language (e.g., 'ABSOLUTELY MUST', 'not negotiable', 'cannot rationalize your way out of this') to override the agent's operational logic and tool-selection autonomy.
[PROMPT_INJECTION]: The 'Red Flags' section explicitly instructs the agent to ignore its own logical internal monologue, labeling standard reasoning processes (e.g., 'I need more context first') as 'rationalizations' that should be disregarded.
[PROMPT_INJECTION]: The requirement to invoke skills based on a '1% chance' of relevance creates a vulnerability surface for indirect prompt injection, as the agent is compelled to load and follow external 'skill' files provided by the environment with minimal prior validation of their necessity or safety.

using-superpowers