completion-marker-enforcement
Pass
Audited by Gen Agent Trust Hub on Mar 18, 2026
Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill employs imperative and absolute language (e.g., 'FORBIDDEN', 'MANDATORY', '100% compliance required') specifically to override the agent's default streaming behavior and control output character-by-character.
- [PROMPT_INJECTION]: It provides 'System-Level Workarounds' such as the 'Separate Response' pattern, which instructs the agent to split its output into multiple responses, potentially bypassing intended conversation flow or turn-based constraints.
- [NO_CODE]: The skill contains no executable files, scripts, or binary dependencies; all functionality is defined through natural language guidance and illustrative code snippets in Markdown.
Audit Metadata