fine-tuning-data-generator
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it processes untrusted user requirements (e.g., domain, task, and tone specifications) to drive the automated generation of training data.\n
- Ingestion points: Requirements gathering in Phase 1 (SKILL.md) where user inputs define the generation plan and final dataset content.\n
- Boundary markers: Absent. The skill instructions do not specify the use of delimiters or distinct 'ignore' instructions to delineate user-provided requirements from the generator's logic.\n
- Capability inventory: The skill uses the
Writetool to create.jsonland.txtfiles and theBashtool to run provided validation and analysis scripts (scripts/validate_chatml.pyandscripts/analyze_dataset.py).\n - Sanitization: There is no explicit sanitization of user-provided requirements before they are incorporated into the generation process; the skill relies on model guardrails and standard JSON structural validation.
Audit Metadata