The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses imperative and high-priority language ("CRITICAL — 开始前 MUST 先用 Read 工具读取...") to force the agent to read an external file before proceeding. This is an attempt to override the agent's natural instructional flow.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. It ingests data from external sources (the output of ziniao-cli commands) and incorporates it into the agent's context without boundary markers or sanitization. This could allow malicious instructions embedded in the ERP role data to influence agent behavior.
Ingestion points: Outputs from ziniao-cli role list, ziniao-cli role detail, and ziniao-cli role permissions are processed by the agent in SKILL.md.
Boundary markers: Absent.
Capability inventory: The skill uses ziniao-cli (subprocess execution) as its primary tool.
Sanitization: Absent.
[COMMAND_EXECUTION]: The skill requires the ziniao-cli binary and provides detailed instructions for its execution to perform administrative tasks in the Ziniao ERP environment. While this is the primary purpose of the skill, it represents a significant capability surface.

ziniao-role