improve-prompt
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied prompts to generate improved instructions, creating a surface for indirect prompt injection.\n
- Ingestion points: User input is ingested via the
$ARGUMENTSvariable inSKILL.md.\n - Boundary markers: There are no explicit boundary markers or instructions to isolate the user-provided text from the skill's operational logic during processing.\n
- Capability inventory: The skill uses
AskUserQuestionand can generate dynamic HTML content.\n - Sanitization: No explicit sanitization or escaping mechanisms are described for user-provided prompt components.\n- [DYNAMIC_EXECUTION]: The skill includes an 'Optional prompt tuner' feature that generates HTML and potentially JavaScript at runtime.\n
- Evidence:
SKILL.mdinstructs the agent to 'create a throwaway HTML prompt tuner' for editing and previews.\n - Impact: This represents runtime script generation. While intended for UI assistance, generating executable web content based on user-influenced variables is a dynamic execution pattern.
Audit Metadata