deepworkplan

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 1.00). These URLs link to remote installer scripts (plain .sh/.ps1 endpoints) that are commonly piped to a shell/PowerShell—an unsafe pattern and a high-risk malware distribution vector unless checksums/trust/verification are explicitly performed.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The repo intentionally includes mechanisms that persist and expose host credentials into containerized agent environments, recommends piping remote install scripts, and ships one-word AI-CLI wrappers that offer explicit “--dangerously” bypass flags — patterns that enable credential access/persistence and bypassing of approval/sandbox controls (high-risk backdoor/exfiltration capability).

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

• Potentially malicious external URL detected (high risk: 0.90). The addon and devcontainer templates include installer commands that fetch-and-execute remote scripts (git clone / curl | bash / irm | iex) at runtime — e.g., https://cli.dailybot.com/install.sh, https://cli.dailybot.com/install.ps1, https://github.com/DailybotHQ/agent-skill.git, https://claude.ai/install.sh, and https://cursor.com/install — which are invoked by the addon flows or container builds and therefore execute remote code.