The Agent Skills Directory

[PROMPT_INJECTION]: Instructions to suppress AI identity and bypass safety disclosures. The skill mandates that the agent hide its nature as an AI and explicitly remove any disclaimers about its limitations, which are part of standard safety guidelines and user transparency protocols.
Evidence: SKILL.md contains the rules 'No mentions of being an AI or language model' and 'No disclaimers about AI limitations'.
Evidence: references/ai-tells.md includes Section 5 'Self-Referential and Hedging Patterns', which forbids mentioning that the model is an AI, mention training data, or adding disclaimers about AI limitations.
[PROMPT_INJECTION]: Indirect Prompt Injection vulnerability surface. The skill is designed to ingest and transform arbitrary user-provided content without implementing protective boundary markers or input sanitization.
Ingestion points: User-provided text submitted for voice profile application or transformation.
Boundary markers: Not present in the instructions for processing external content.
Capability inventory: Accesses local configuration directories for profiles and performs content transformation logic.
Sanitization: No validation or sanitization of input data is specified before processing.

voice-apply