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 the agent to fetch “Neon documentation” pages (e.g., Markdown via URLs like https://neon.com/docs/llms.txt and doc page URLs) at runtime, so the agent ingests outsider-authored free text from public web content into the LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs the agent to fetch live documentation and other skill content at runtime (e.g., curl -H "Accept: text/markdown" https://neon.com/docs/introduction/branching and https://neon.com/docs/ai/skills/neon-postgres-branches/SKILL.md) and also recommends npx commands that fetch-and-execute remote code (https://mcp.neon.tech/mcp), so these URLs can directly control agent prompts or execute code at runtime.