bailian-opd
Pass
Audited by Gen Agent Trust Hub on Nov 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied dataset files for model training, which represents a potential surface for indirect prompt injection if those files contain malicious instructions.\n
- Ingestion points: Training and validation datasets provided via the
--datasetsand--validationsflags inscripts/opd-preflight.mjs.\n - Boundary markers: No explicit markers or delimiters are used in the scripts to isolate file content from instructions.\n
- Capability inventory: The skill executes commands via the
blCLI tool to validate datasets and manage training jobs.\n - Sanitization: The skill performs local schema validation using
bl dataset validate --schema chatmlbefore job submission to ensure data format integrity.\n- [COMMAND_EXECUTION]: The skill scripts execute theblCLI tool usingnode:child_process.spawnSyncto perform various cloud operations.\n - Execution logic: The implementation passes arguments as an array and disables shell execution (
shell: false), which follows security best practices to prevent command injection.\n - Safety gates: Critical actions such as job submission are protected by a manual consent gate, requiring an explicit
--confirmedflag from the user after they have reviewed a dry-run report.
Audit Metadata