mcp-server-scaffolding-asyncio

Pass

Audited by Gen Agent Trust Hub on Mar 31, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill provides a template for an MCP resource provider that is vulnerable to indirect prompt injection and path traversal.
  • Ingestion points: The read_resource function in SKILL.md accepts a uri string from the agent's context, which could be influenced by external data.
  • Boundary markers: Absent in the provided template.
  • Capability inventory: The template implementation uses aiofiles.open to perform file system read operations based on the provided URI.
  • Sanitization: The code only checks if the URI starts with file:/// and fails to perform path normalization or validation to ensure the resulting path is restricted to an intended directory, potentially allowing an attacker to read arbitrary files via traversal patterns (e.g., ../../etc/passwd).
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 31, 2026, 09:08 AM
Security Audit — agent-trust-hub — mcp-server-scaffolding-asyncio