wuji-cli-logs
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
wujicommand-line utility for log and diagnostic operations. - [DATA_EXFILTRATION]: The skill gathers diagnostic data from
~/.wuji/logs/and~/.wuji/calibration/recordings/to generate support bundles. Although it implements a multi-layered redaction strategy for credentials and PII, the exported ZIP files contain detailed technical metadata about the device and environment. - [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing external log files that may contain adversarial instructions. (1) Ingestion points: Log files (
sdk_*.log,studio_*.log) and device diagnostics. (2) Boundary markers: No delimiters are specified to isolate log data from instructions. (3) Capability inventory: Filesystem access and bundle creation viawujiCLI. (4) Sanitization: Credentials are redacted, but log content is not filtered for malicious instructions.
Audit Metadata