kit-author

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). Several entries are unpinned or malformed GitHub refs (branches, tags, empty ref/dir) and two generic install scripts (example.com and myservice.com) point to remote shell installers — patterns that can be used to distribute malware; official docs and pinned-SHA refs (and code-server.dev / claude.ai examples) are lower risk and not included.

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

• Potentially malicious external URL detected (high risk: 0.90). The docs contain runtime fetch-and-exec commands (e.g., curl -fsSL https://code-server.dev/install.sh | sh and curl -fsSL https://myservice.com/install.sh | bash, also curl -fsSL https://claude.ai/install.sh | bash) and pinned git+https kit references (git+https://github.com/org/repo.git#ref=&dir=auditor and git+https://github.com/org/repo.git#ref=<40-hex-sha>&dir=) which the engine runs or clones at sandbox creation and can therefore execute remote code or inject remote spec content that controls agent prompts/instructions.