skills/modelstudioai/cli/bailian-gen/Gen Agent Trust Hub

bailian-gen

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute the bl CLI tool for interacting with Aliyun Bailian services. It also suggests using ffmpeg for local video post-processing tasks like concatenation and audio muxing.
  • [INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection as it processes untrusted user data (images, videos, and audio) through multimodal models (bl vision, bl omni).
  • Ingestion points: bl image edit, bl video edit, bl omni, bl vision describe, and bl speech recognize accept user-provided file paths or URLs.
  • Capability inventory: The skill has the capability to write to the local file system (--out, --download) and execute shell commands (bl).
  • Boundary markers: None explicitly mentioned for model inputs.
  • Sanitization: Not specified for model outputs before they are processed by the agent.
  • [DATA_HANDLING]: The bl CLI automatically handles the upload of local files to DashScope temporary storage (OSS) for processing. This is a standard functional requirement for the service and is clearly documented.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 08:56 AM
Security Audit — agent-trust-hub — bailian-gen