alibabacloud-sls-data-agent-qa
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Python script (scripts/context_model_qa.py) to interact with Alibaba Cloud services. The script is well-hardened, using argparse for safe argument handling and validating endpoint URLs to ensure they use HTTPS and contain no malicious path or credential components.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external logs which could contain untrusted content. 1. Ingestion points: Log data retrieved via query_logs and model metadata via ContextModelClient in scripts/context_model_qa.py. 2. Boundary markers: Instructions in SKILL.md emphasize result validation and cite-only evidence. 3. Capability inventory: Network access to Alibaba Cloud services and shell execution of the provided Python scripts. 4. Sanitization: The script includes literal() for escaping SPL queries and an error_payload() function that redacts environment variables from error outputs to prevent credential leakage.
- [SAFE]: All external dependencies are official Alibaba Cloud SDK packages. The script follows best practices for authentication by utilizing the standard alibabacloud-credentials provider, avoiding hardcoded secrets.
Audit Metadata