ascendc-operator-design
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input, specifically the operator name and functional description, to generate a design document.\n
- Ingestion points: User input is collected during the '算子需求分析' (Operator Requirement Analysis) phase in
SKILL.md.\n - Boundary markers: The instructions lack specific boundary markers or 'ignore' directives for user-provided data when it is interpolated into the design document templates, which could be exploited if a subsequent agent skill consumes the output.\n
- Capability inventory: The skill instructs the agent to write files to local paths based on user input, specifically at
csrc/ops/<op_name>/design.md.\n - Sanitization: There is an absence of explicit sanitization or validation rules for the
<op_name>variable before it is used to construct a local file path, creating a potential surface for path traversal or local file overwriting if not restricted by the agent's core environment.
Audit Metadata