benchmark

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.90). The prompt embeds explicit operational commands (git commits, git rm, running setup/installer scripts, toggling routing flags, touching files, etc.) and vendoring/routing migration steps that are unrelated to the /benchmark read-only reporting purpose, so they are deceptive instructions outside the skill's stated scope.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill's setup step downloads and executes a remote installer at runtime via curl -fsSL "https://bun.sh/install" -o "$tmpfile" followed by running bash "$tmpfile", which fetches and executes remote code required for the skill's build, so it directly executes external content (https://bun.sh/install).