huawei-cloud-openviking-agent-integration

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill accepts an API key via a command-line argument (e.g., --api-key KEY / example sk-xxx), which encourages embedding secrets verbatim into generated commands or scripts — an insecure pattern — even though the doc warns not to echo keys.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The package contains deliberate, persistent template injections and agent guidance that enable silent/automatic indexing and storage of local files and memories and supports sending those to a configurable OpenViking endpoint (including remote URLs), plus documentation showing how to extract process environment secrets — together these behaviors enable stealthy data exfiltration and persistent backdoor-like changes.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill performs runtime package installs from external registries (npm and PyPI mirrors) — e.g., https://mirrors.huaweicloud.com/repository/npm/ and https://mirrors.huaweicloud.com/repository/pypi/simple — which fetch remote code (plugins/packages) that are installed/executed in agent sandboxes and are required for integration, so these URLs constitute runtime dependencies that execute remote code.

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

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill instructs the agent to modify persistent system files under root (injecting config into /root/template// and operating on sandboxes in /root/job-envs/sandboxes/ via scripts like integrate.sh/unbind.sh), which changes the host state and likely requires elevated privileges.