bailian-finetune

Pass

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill facilitates the use of the bl binary for model-related tasks, including dataset management, training job creation, and service deployment. Commands such as bl dataset upload and bl finetune create are central to the skill's functionality.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes user-provided dataset files in .jsonl and .zip formats, creating a potential vector for indirect prompt injection if malicious content is included in the data.\n
  • Ingestion points: Dataset files are read during local validation (bl dataset validate) and upload (bl dataset upload) processes, as well as during training job initiation.\n
  • Boundary markers: The instructions do not define specific delimiters or guidelines to help the agent differentiate between data content and instructional commands when reviewing dataset summaries or error logs.\n
  • Capability inventory: The skill possesses the ability to delete datasets, jobs, and deployments, and can initiate training tasks that utilize cloud resources/quota.\n
  • Sanitization: Validation is performed structurally by the CLI tool, but no specific sanitization is mentioned for the agent's interaction with the data content itself.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 22, 2026, 12:53 PM
Security Audit — agent-trust-hub — bailian-finetune