bailian-managed-agent
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill acts as a wrapper for the
bl managed-agenttoolset, allowing the agent to perform a wide range of infrastructure-as-code (IaC) tasks. It manages resources such as agents, deployments, memory stores, and webhooks via command-line execution.\n- [EXTERNAL_DOWNLOADS]: The skill facilitates the retrieval of remote resources including agent skill definitions and files from the Alibaba Cloud Bailian service. These operations are essential to its function as a management tool for cloud-hosted agents.\n- [INDIRECT_PROMPT_INJECTION]: The skill handles untrusted data from external sources.\n - Ingestion points: Configuration files (agents.yaml), local secret stores (vault.json), and remote session event history.\n
- Boundary markers: The instructions mandate a 'plan-then-confirm' workflow, requiring the agent to display the diff to the user before applying any state-changing operations.\n
- Capability inventory: Subprocess execution of the 'bl' CLI to perform network operations, file management, and resource mutation.\n
- Sanitization: Relies on user verification of the generated plan as a gate for validation.
Audit Metadata