The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-supplied lists of ideas and initiatives via the $ARGUMENTS variable and additional template placeholders. This creates a surface for indirect prompt injection, as an attacker could insert instructions within the data to manipulate the model's behavior.
Ingestion points: The $ARGUMENTS variable and the {{WHAT ARE YOU OPTIMIZING FOR?}} placeholder in SKILL.md.
Boundary markers: The template uses <context> tags to isolate the user list.
Capability inventory: The skill includes no code, shell commands, or network-capable tools, which restricts the potential impact of any successful injection.
Sanitization: There are no explicit instructions for the model to sanitize or ignore instructions embedded within the user data.
[NO_CODE]: The skill consists exclusively of markdown instructions and prompt templates, with no associated scripts or executables.

prioritize