liki-divination
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the
jsonschemalibrary from a standard package registry if it is not already present on the system. - [COMMAND_EXECUTION]: The skill utilizes a local Python CLI (
tools/agent_cli.py) to execute its core logic. This CLI acts as a dispatcher for divination functions, using structured JSON inputs and outputs. - [DATA_EXFILTRATION]: The skill transmits divination parameters and diagnostic information (agent metadata, session hashes, and error summaries) to the vendor's API endpoint at
liki.hk. This is standard functionality for the service's operation and feedback mechanism. - [INDIRECT_PROMPT_INJECTION]: The skill processes user questions which are interpolated into tool parameters. To mitigate risks, it implements a robust safety assessment layer (
tools/divination_safety.py) that filters high-risk topics like self-harm or medical emergencies, and it employs strict schema validation for all tool responses.
Audit Metadata