lead-scoring-model

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data which could contain malicious instructions designed to influence the agent's behavior. \n
  • Ingestion points: Workflow Step 1 in SKILL.md and the Inputs section in references/inputs.md specify loading CRM exports, historical win/loss data (CSV), and sales activity logs. \n
  • Boundary markers: The instructions do not define delimiters or specific 'ignore embedded instructions' warnings for the data being analyzed. \n
  • Capability inventory: The skill is granted Bash, Write, WebFetch, and WebSearch capabilities, allowing it to execute code and perform network operations. \n
  • Sanitization: There is no mention of sanitizing, escaping, or validating the content of the CRM exports before the agent processes them using the Bash tool. \n- [COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to perform data audit, win/loss pattern analysis, and model validation tasks. The workflow involves executing steps that likely require the generation and execution of scripts (e.g., Python) to calculate statistical metrics like F1 scores and AUC-ROC. While these operations are functional requirements for a data scientist agent, they represent a significant capability tier that could be abused if the agent is manipulated via malicious input data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 01:02 PM
Security Audit — agent-trust-hub — lead-scoring-model