skills/ml8s/liki/liki-fengshui/Gen Agent Trust Hub

liki-fengshui

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill communicates with external endpoints at https://liki.hk to perform version checks and retrieve JSON-RPC schemas during initialization.
  • [DATA_EXFILTRATION]: User-provided data, including birth year, gender, and specific home orientation details, is transmitted to the remote endpoint https://liki.hk/jsonrpc to generate Feng Shui charts. Feedback is also sent to https://liki.hk/api/feedback.
  • [COMMAND_EXECUTION]: The skill logic includes a procedure to compare a local VERSION file with a remote version and instructs the agent to prompt the user with update commands if a mismatch is found, which could lead to the execution of unverified shell commands.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data returned from the liki.hk JSON-RPC service, which represents a potential injection surface.
  • Ingestion points: Data is ingested via the JSON-RPC responses from methods such as bazhai.chart, xuankong.chart, and time.now.
  • Boundary markers: The skill defines strict boundaries, instructing the LLM to only interpret "returned factors" and forbidding the manual modification or assembly of the chart_digest returned by the engine.
  • Capability inventory: The skill maintains network access to the liki.hk domain for RPC calls, versioning, and feedback.
  • Sanitization: The skill utilizes the rpc.discover method to identify and adhere to the schema provided by the engine before processing data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 05:02 PM
Security Audit — agent-trust-hub — liki-fengshui