greptimedb-quickstart

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). All listed URLs are legitimate documentation, vendor pages, or local endpoints except for a direct raw.githubusercontent.com link to an install.sh script which is a higher-risk pattern (downloading a shell script and piping to sh is a common malware distribution vector).

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 sister skills and docs at runtime (e.g., “fetch its SKILL.md URL” and “fetch <https://docs.greptime.com/llms.txt> / .md pages”), which are outsider-authored free-form markdown that can be ingested into the LLM context via those fetched URLs.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs the agent to fetch remote skill documents and machine-readable docs at runtime (e.g. https://docs.greptime.com/skills/greptimedb-pipeline/SKILL.md and https://docs.greptime.com/llms.txt) and also includes a curl | sh install command that executes remote code (https://raw.githubusercontent.com/greptimeteam/greptimedb/main/scripts/install.sh), so remote content can control agent behavior or run code.