claude-code-best

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.85). These URLs mix legitimate services (api.anthropic.com, bun.sh, cloud.langfuse.com) with untrusted/odd domains and an unfamiliar GitHub user (remote-control.claude-code-best.win, ccb.agent-aura.top, github.com/claude-code-best/...) that request remote-control, filesystem access, and custom runtime installs—making the bundle suspicious for potential malware/remote-access abuse.

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime can ingest outsider-authored free text via its web search integration (e.g., fetching/searching the web and then passing retrieved page/snippet text into the agent context), which is explicitly enabled when search API keys are set.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill documents a runtime "remote control" feature that connects to an external server (CLAUDE_BRIDGE_BASE_URL=https://remote-control.claude-code-best.win/) which can push messages/commands into the agent session (directly controlling prompts), and the docs also show install/run steps that execute remote code (e.g., curl -fsSL https://bun.sh/install | bash, git clone https://github.com/claude-code-best/claude-code.git, and npx @modelcontextprotocol/create-server) — all of which are runtime/setup fetch-and-execute points of control.