power-bi-business-analysis

Pass

Audited by Gen Agent Trust Hub on Apr 20, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill facilitates potential indirect prompt injection by processing raw stakeholder input for documentation purposes.\n
  • Ingestion points: The stakeholder-interview-template.md reference file directs the agent to capture user responses verbatim.\n
  • Boundary markers: There are no specific delimiters or instructions to treat stakeholder input as untrusted data.\n
  • Capability inventory: The skill has shell execution capability (python validation) and its outputs drive downstream development skills with additional privileges.\n
  • Sanitization: No sanitization or validation of stakeholder-provided text is performed before it is used in documentation.\n- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local validation script via the shell.\n
  • Evidence: SKILL.md contains an instruction to run 'python agents/validate_handoff.py' to verify the integrity of handoff JSON files. This is a project-specific vendor resource.\n- [EXTERNAL_DOWNLOADS]: The skill references and fetches external technical documentation.\n
  • Evidence: SKILL.md directs the agent to use 'microsoft-learn-mcp' tools to search and fetch content from Microsoft Learn. This is a well-known and trusted technical information service.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 20, 2026, 04:21 AM