agent-architecture

Pass

Audited by Gen Agent Trust Hub on May 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill's installation command downloads the instruction content (SKILL.md) from a public GitHub repository (ollieb89/viflo).
  • [DATA_EXFILTRATION]: The skill provides example code for a fetch_url tool that uses urllib.request (Python) and fetch (TypeScript) to retrieve content from arbitrary URLs. This capability creates a risk of Server-Side Request Forgery (SSRF), where an agent could be manipulated into accessing sensitive internal network services or cloud metadata endpoints (e.g., 169.254.169.254).
  • [PROMPT_INJECTION]: The inclusion of web retrieval functionality creates a surface for indirect prompt injection. If the agent fetches a malicious website, the content could contain hidden instructions that override the agent's primary task.
  • Ingestion points: Content retrieved via the fetch_url tool in SKILL.md.
  • Boundary markers: None present in the provided code examples to separate external data from system instructions.
  • Capability inventory: The agent is granted network access via HTTP clients (urllib.request, fetch).
  • Sanitization: No sanitization or validation of the fetched content is performed before it is added to the model's context.
Audit Metadata
Risk Level
SAFE
Analyzed
May 14, 2026, 06:26 AM
Security Audit — agent-trust-hub — agent-architecture