neon-postgres

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). SKILL.md instructs fetching Neon documentation pages as Markdown (public web content) at runtime, which would be outsider-authored free text ingested into the agent’s LLM context via the docs-fetching workflow.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs fetching Neon documentation at runtime and using it as the "source of truth" for responses (e.g., https://neon.com/docs/introduction/branching.md), which injects remote content into the agent context and thus can directly influence prompts.