check-training-data
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and audit untrusted datasets (e.g., web text, chat transcripts) and utilizes LLMs to judge or score this data.
- Ingestion points: Untrusted data enters the context via
pb.Validate(data=...)inSKILL.md. - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are mentioned in the
prompt()orcalibrate_prompt()methods. - Capability inventory: The skill can perform network operations (API calls to Anthropic/Cohere), write files (dataset cards, reports), and execute local code (ablation evaluation).
- Sanitization: There is no evidence of input sanitization or escaping of the processed text before it is sent to the LLM judge.
- [DYNAMIC_EXECUTION]: The skill allows the definition and execution of arbitrary code for scorers and evaluation functions.
- Evidence:
pb.Ablationandpb.Validatetake user-provided Python functions (evaluate,scorer) which are executed during the interrogation and ablation process. - Evidence:
pb.yaml_ablation()supports anallow_python=Trueflag that imports and executes code defined in a YAML file'sevaluate:block. - [EXTERNAL_DOWNLOADS]: The skill relies on and interacts with well-known external services to download models and perform remote inference.
- Evidence: Utilizes the
transformersandsentence-transformerslibraries to download and cache models from Hugging Face. - Evidence: Loads specific third-party models for language identification (e.g.,
lid.176.binfor fastText) and perplexity (kenlm). - Evidence: Performs remote API calls to services like Cohere (via
pb.scorers.reranker) and Anthropic (via.prompt()).
Audit Metadata