assess
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill demonstrates safe behavior, focusing on its primary purpose of tax assessment through structured dialogue and local state management. No obfuscation, unauthorized command execution, or persistence mechanisms were detected.- [DATA_EXFILTRATION]: The skill accesses local configuration files and tax records (e.g.,
shinkoku.config.yamland past tax returns) to provide accurate assessments. However, no network operations or external exfiltration attempts were found; all data remains within the local environment.- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources and user input with the following security profile: - Ingestion points: User input, local configuration files, and step-based progress files (
.shinkoku/progress/). - Boundary markers: Absent; the skill relies on direct logic flows.
- Capability inventory: Limited to reading and writing local files using the agent's provided tools.
- Sanitization: Not explicitly implemented, but the skill interprets data strictly within the context of tax law evaluation logic.
Audit Metadata