The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from the user (such as product names and audience fears) to generate marketing copy, which could be exploited to influence the agent's output.
Ingestion points: Input schema defined in SKILL.md for product.name, your_bonuses, and audience_fears fields.
Boundary markers: The skill includes a 'Self-Validation' checklist and a 'Quality Gate' to guide the agent in verifying the output.
Capability inventory: No dangerous capabilities detected; the skill lacks network access, file writing, or command execution features.
Sanitization: No automated sanitization is present; the skill relies on instructional validation rules.

guarantee-generator