attention-kind
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The instructions utilize assertive and imperative language designed to prioritize the skill's persona over other instructions. Phrases such as 'Read that twice, it matters more than any rule below' and 'nothing below overrides it' are identified as patterns used to force compliance and potentially override default agent behaviors.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and reformat user-provided text or conversation history into an 'ADHD-friendly' style. This creates a vulnerability surface where malicious instructions embedded in the input text could influence the agent's behavior.
- Ingestion points: The skill processes the 'rest of the conversation' and specific user-provided deliverables like emails or code snippets for reformatting.
- Boundary markers: The instructions do not define boundary markers or delimiters for untrusted user content.
- Capability inventory: The skill's primary capability is text formatting, tone adjustment, and code comment modification.
- Sanitization: There are no explicit instructions for the agent to sanitize or escape external content before processing it for delivery.
Audit Metadata