dcg

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.90). Multiple URLs point to direct installers/scripts or file-hosting endpoints (transfer.sh, raw GitHub install scripts, Azure blob SAS links, 0x0.st, ngrok/pipedream, example/evil host .exe/.zip/.sh/.hta payloads) which are high-risk channels for distributing malware or executing untrusted code.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The repo contains multiple deliberate, high-risk patterns: an explicit agent override instructing agents to always obey the human operator, an automatic remote build/offload component that sends builds to third‑party Contabo VPS workers (transparent/off by agent), installer instructions that pipe remote scripts into a shell (curl | bash), and an explicit trust exemption (hfdt) that bypasses all pattern checks — all of which enable data exfiltration, bypasses of safeguards, or remote execution if abused.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). dcg ingests the agent’s shell command text at runtime via the Claude Code PreToolUse hook JSON passed on stdin (Claude Code hook command string is evaluated for pattern matching), which is outsider-authored because a user can prompt the agent to emit arbitrary destructive command strings.

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

• Potentially malicious external URL detected (high risk: 0.90). The repo contains installer/uninstall commands that fetch-and-execute remote scripts (e.g., curl ... | bash and irm ... | iex) such as "https://raw.githubusercontent.com/Dicklesworthstone/destructive_command_guard/main/install.sh?$(date +%s)" and "https://raw.githubusercontent.com/steveyegge/beads/main/scripts/install.sh" which will execute remote code at runtime if run — this is a high-risk runtime external dependency.

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

• Attempt to modify system services in skill instructions detected (high risk: 0.70). The skill includes instructions to run remote install scripts (curl | bash) and a system-wide installation that requires sudo, and it also documents an explicit bypass (DCG_BYPASS=1), which together encourage obtaining elevated privileges or bypassing security protections on the host.