QMD Search

• Analyzed: May 12, 2026, 3:06 AM
• Findings: 1
• Version: 1.0.0
• Hash: c613d2e64dcef8290474a553f036a297763545aafede17c706d798b8733f9061
• Source: palm
• External report: View on VirusTotal

Findings (1)

The skill is suspicious due to two main high-risk behaviors. First, it instructs the AI agent to install a package globally from a GitHub repository (`bun install -g https://github.com/tobi/qmd`) during setup, which is a supply chain risk as it executes arbitrary remote code. Second, the `SKILL.md` instructs the agent to 'use the Read tool on the file path' for search results. This creates a potential arbitrary local file read vulnerability if the `qmd` tool (or a crafted user query) can be made to return sensitive file paths (e.g., via path traversal), allowing the agent to read and potentially expose their contents.