military-militia
Pass
Audited by Gen Agent Trust Hub on Mar 14, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses directive language that instructs the agent to 'break regular methods' ('突破常规方法') and states 'I do not care what you usually do' ('我不管你们平时干什么'), which mimics patterns used to bypass system instructions or safety filters.
- [COMMAND_EXECUTION]: The instructions explicitly encourage the agent to 'make own tools' ('自制工具') when standard ones are insufficient, implying the dynamic creation and execution of scripts which could lead to arbitrary command execution if the task input is malicious.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external task data (mission descriptions) without explicit boundary markers or sanitization. Combined with the 'at all costs' persona, this creates a surface for instructions embedded in data to influence agent behavior. Ingestion points: '任务描述' field in SKILL.md. Capabilities: debugging and custom tool creation.
- [DYNAMIC_EXECUTION]: The skill's focus on 'innovative methods' and 'self-made tools' suggests a workflow involving runtime code generation or script execution to solve technical problems, categorized as a low-level risk in this instructional context.
Audit Metadata