aminer-deep-search

Pass

Audited by Gen Agent Trust Hub on Jul 24, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a ReAct-style agent for academic literature reviews. Technical review of the code shows standard patterns for API interaction and multi-threading without any malicious indicators.
  • [DATA_EXFILTRATION]: Outbound network requests are limited to the official AMiner API gateway and the user-specified LLM endpoint. Sensitive credentials like AMINER_API_KEY and LLM_API_KEY are managed through environment variables and are protected from exposure in logs or outputs.
  • [COMMAND_EXECUTION]: The execution flow involves a local Python script (react_agent.py) triggered via the CLI. The skill logic does not utilize unsafe execution functions like eval or exec, and it relies on standard argument parsing for input handling.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes paper metadata (titles and abstracts) retrieved from AMiner. This data enters the agent's context and influences decision-making. However, the risk is mitigated by the use of structured JSON boundaries and the restriction of tool capabilities to academic search functions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 24, 2026, 08:32 AM
Security Audit — agent-trust-hub — aminer-deep-search