icp-contabilidade-beatriz-exemplo
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, command execution, or data exfiltration attempts were found during the analysis of the skill's instructions and data files.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process specific external context (profile, journey, and evidence data) to personify a character. While this creates a surface for potential indirect prompt injection from the processed data, the instructions include specific safeguards to maintain simulation boundaries, such as identifying responses as simulations and verifying claims against evidence, resulting in a low risk typical for context-driven agents.
- [EXTERNAL_DOWNLOADS]: The skill references multiple external URLs in its methodology documentation (
fontes.md). These links point to well-known, authoritative marketing and UX research organizations and are used exclusively for professional reference and educational purposes rather than code execution or package installation. - [NO_CODE]: The skill contains only Markdown instructions and structured data files (JSON/YAML). No executable scripts, binaries, or automated package installations are present, significantly reducing the attack surface.
Audit Metadata