vpe-review

Pass

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell commands to run local Python scripts (delivery_throughput_analyzer.py, eng_hiring_funnel_calculator.py, and eng_team_structure_designer.py) located in a relative directory structure (../../../skills/vpe-advisor/scripts/). These scripts are used to process metrics and hiring data.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external files without explicit sanitization or boundary markers.
  • Ingestion points: sprint_metrics.json, funnel.json, and team.json files read via the workflow section.
  • Boundary markers: No delimiters or instructions are used to isolate the data from the prompt context.
  • Capability inventory: Execution of local Python scripts with file arguments.
  • Sanitization: No evidence of validation or filtering for the input JSON files.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 18, 2026, 04:11 PM
Security Audit — agent-trust-hub — vpe-review