link-reader

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFE
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill handles sensitive API keys by checking environment variables and providing instructions for the agent to store them in a workspace file. This is part of the vendor's documented integration flow.\n- [PROMPT_INJECTION]: The skill processes data from external URLs, which represents a surface for indirect prompt injection.\n
  • Ingestion points: External content enters the system through scripts/link_reader.py from user-provided URLs.\n
  • Boundary markers: There are no specific delimiters or instructions to the LLM to ignore commands within the fetched content.\n
  • Capability inventory: The skill uses the volcenginesdkarkruntime library to make network calls to the vendor's tool execution API.\n
  • Sanitization: The raw text and titles from the URLs are returned to the agent without any sanitization of the content for malicious instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 12:24 AM