loom-model-evaluation
Pass
Audited by Gen Agent Trust Hub on Jul 25, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill is designed to ingest and evaluate external data, including LLM-generated responses and production datasets, without explicit sanitization or boundary markers. This creates a surface for indirect prompt injection where instructions embedded in the evaluated data could influence the agent's behavior.
- Ingestion points: The
evaluate_llm_generationfunction in Example 6 and theModelMonitorclass in Example 8 process external content. - Boundary markers: No delimiters or instructions to ignore embedded commands are specified.
- Capability inventory: The skill frontmatter allows the use of
Bash,Write,Edit, andReadtools. - Sanitization: There is no evidence of filtering or validation for the data being ingested for evaluation.
- [COMMAND_EXECUTION]: The skill requests access to the
Bashtool in its frontmatter, enabling the execution of shell commands and scripts required for model evaluation workflows.
Audit Metadata