skills/cascade-protocol/agentbox/web/Gen Agent Trust Hub

web

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data from arbitrary web URLs, creating a surface for indirect prompt injection attacks where malicious instructions on a webpage could influence the agent's behavior.
  • Ingestion points: External content is retrieved and processed through the /v1/crawl and /v1/search endpoints.
  • Boundary markers: The documentation does not specify the use of delimiters or guardrail instructions to separate retrieved web content from system instructions.
  • Capability inventory: While the skill contains no local scripts, it uses the x402_payment tool to interact with the Solana blockchain.
  • Sanitization: There is no mention of sanitization or filtering of the retrieved content before it is passed to the agent.
  • [DATA_EXFILTRATION]: The skill sends data (URLs and search queries) to the external domain web.surf.cascade.fyi. This behavior is the primary intended function of the skill and targets the vendor's own infrastructure.
  • [NO_CODE]: The skill does not contain any local executable scripts or binary files, relying entirely on tool calls to an external API.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 07:54 PM