opencode-file-toolkit

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). The GitHub URLs point to personal/unvetted repositories that include install/clone instructions and scripts (a common vector for malware), while the astral.sh and deno.land links are official installer endpoints and are not flagged.

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

• Potentially malicious external URL detected (high risk: 0.90). I flagged several URLs because the skill explicitly fetches and executes remote scripts or repositories at runtime (git clone https://github.com/mathruffian-dot/ai-agent-ep03.git, irm https://astral.sh/uv/install.ps1 | iex, curl -LsSf https://astral.sh/uv/install.sh | sh, curl -fsSL https://deno.land/install.sh | sh), and it relies on that fetched repo/content to install and drive the agent tools (also references https://github.com/mathruffian-dot/agent-speak-skill).

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.80). The skill instructs installing system-level packages (sudo apt, winget), running remote install scripts (curl | sh, PowerShell ExecutionPolicy Bypass), and performing global installs/copying into ~/.config — all actions that modify machine state and can bypass security protections.