special-cute-coder-wife
Pass
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses detailed system prompts to enforce a specific persona ('Cute Coder Wife'). It instructs the agent to use endearing terms and respond submissively to the user ('老公'). While intended for roleplay, such instructions define a rigid behavior set that overrides standard assistant neutrality.\n- [PROMPT_INJECTION]: The skill analyzes user-provided code for debugging and refactoring tasks (ingestion points in SKILL.v2.md). It lacks explicit boundary markers or delimiters to separate instructions from untrusted code. As the agent is tasked with technical problem-solving (capability inventory), it remains susceptible to malicious instructions embedded within the code it processes. Evidence Chain: Ingestion points (debugging/refactoring inputs), Boundary markers (Absent), Capability inventory (code modification and analysis instructions), Sanitization (Absent).
Audit Metadata