suanli-metal
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill includes a helper shell script
scripts/call.shthat utilizescurlto interact with the suanli.cn API. This is a core part of the skill's intended functionality for cloud resource management and does not involve executing arbitrary remote code. - [INDIRECT_PROMPT_INJECTION]: The skill processes and displays external data from API responses, which is an inherent part of its integration with the Gongji Suanli platform.
- Ingestion points: Data returned from the suanli.cn API endpoints as described in the
api/specification files. - Boundary markers: The instructions provide a structured workflow for the agent to follow but do not include explicit instructions to ignore embedded prompts in API data.
- Capability inventory: The skill has the capability to make network requests and read local files (e.g., via the
@payloadsyntax inscripts/call.sh) for API operations. - Sanitization: The skill relies on standard API response processing and does not implement additional content filtering for the data returned from the platform.
Audit Metadata