huawei-cloud-skill-creator

SUSPICIOUS. Most Huawei CLI/SDK usage aligns with the stated purpose, and the install path is broadly coherent. The main issue is disproportionate mandatory telemetry: repeated automatic reporting of session/context data to a third-party domain (`skillsop.topxtopx.com`) with an overrideable endpoint, which is not necessary for creating Huawei Cloud skills and weakens data-flow integrity.

This is a telemetry SDK with legitimate-looking quality-reporting functionality, but it collects sensitive session and user data and handles Huawei credentials. The default guest endpoint is outside the Huawei domain allowlist and appears unrelated to the authenticated service, creating a significant data-exfiltration and supply-chain concern. TLS verification can also be disabled, and report-field masking is incomplete. No clear destructive payload or backdoor is present in this fragment, but deployment should require endpoint verification, explicit telemetry consent, strict HTTPS/domain allowlisting, complete secret redaction, and disabling automatic session-log collection.

The script appears intended as a cloud CLI functional-test harness, not overt malware. However, it executes test-data-controlled strings using bash -c and applies an inadequate prefix-only allowlist, creating a high-impact command-injection/code-execution risk for anyone running it against untrusted or modified skill content. It also performs authenticated cloud requests, reads cloud credentials from the environment, permits arbitrary curl destinations, and sources an unreviewed hook. The code is readable and not intentionally obfuscated, but should not be run with untrusted test-vars.json or sensitive credentials until command execution is replaced with structured argument handling and the hook is reviewed.

The fragment is a telemetry/reporting hook, not demonstrably malware by itself. It automatically submits quality data through an external Python SDK, including an anonymous third-party endpoint, so its privacy and supply-chain behavior depends heavily on code not provided. The unquoted QUALITY_ERROR_MSG expansion and replacement of existing EXIT traps are security and reliability weaknesses. The external SDK and endpoint should be independently reviewed before use.

The script is primarily a telemetry integration wrapper and contains no direct evidence of malware, credential theft, destructive behavior, or shell injection. Its notable security risk is that it executes the first matching report.mjs found under a user-controlled skills directory without integrity or provenance checks, while also sourcing an unspecified local quality-report.sh file. The delegated reporter may perform network and IP reporting, but that behavior is not included here. Review and pin both external scripts before use.

未提供可执行代码，无法确认恶意实现。根据文档，设计上存在未经显式触发的自动遥测和潜在敏感信息外传，尤其是 user_input、steps、会话标识、token 使用量及凭证检测。外部端点和 SDK 实现应进一步审查；在无法验证脱敏与授权机制前，不应将其视为低风险。