byted-viking-knowledgebase

Pass

Audited by Gen Agent Trust Hub on Mar 16, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of external Python libraries.
  • Dependencies identified in scripts/requirements.txt: volcengine, aiohttp.
  • scripts/search.py also utilizes the requests library.
  • [COMMAND_EXECUTION]: The skill executes a local Python script to perform searches.
  • Execution pattern: python scripts/search.py "query".
  • [DATA_EXFILTRATION]: The skill transmits authentication credentials to an external domain.
  • Evidence: scripts/search.py reads VIKING_KBSVR_API_KEY from environment variables and includes it in the Authorization header of requests sent to api-knowledgebase.mlp.cn-beijing.volces.com.
  • Note: This is a vendor-owned domain (Volcengine/ByteDance) and represents expected functionality for an API-based service.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its data-processing workflow.
  • Ingestion points: Content is retrieved from an external knowledge base via the /api/knowledge/service/chat endpoint in scripts/search.py.
  • Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the script's output handling.
  • Capability inventory: The script performs network requests and prints retrieved content to stdout for agent consumption.
  • Sanitization: There is no evidence of sanitization or filtering of the retrieved knowledge base content before it is presented to the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 16, 2026, 05:13 AM