langchain-agent

Fail

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS] (HIGH): The skill requires cloning a repository from https://github.com/Eng0AI/langchain-agent.git. This source is not on the Trusted External Sources list, making it an unverifiable dependency.
  • [REMOTE_CODE_EXECUTION] (HIGH): Following the download from an untrusted source, the skill instructs the user to run pnpm install and pnpm dev. This effectively executes the logic contained within the external repository without any verification or sandboxing.
  • [CREDENTIALS_UNSAFE] (SAFE): While the skill mentions sensitive variables like POSTGRES_URL and OPENAI_API_KEY, it uses them as placeholders for user-provided environment variables rather than hardcoding actual credentials.
  • [PROMPT_INJECTION] (LOW): The skill is described as an agent with a 'search' tool. This creates an attack surface for indirect prompt injection, where malicious content retrieved from the web could attempt to influence the agent's behavior at runtime.
  • Ingestion points: Search engine results processed by LangChain.
  • Boundary markers: None specified in the documentation.
  • Capability inventory: Logic execution via LangGraph and tool invocation.
  • Sanitization: Not addressed in the setup instructions.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 17, 2026, 06:10 PM