huawei-cloud-install-openjiuwenswarm

SUSPICIOUS. The skill’s overall purpose is coherent with deployment, but its actual footprint is heavier than a simple local installer: it silently installs dependencies, reads local credentials, downloads and executes an unspecified runtime archive, and persists global commands. The biggest issue is install/execution trust: the runtime mirror source is not verifiably official, yet the downloaded binaries are executed and configured with local credentials.

Moderate-to-high security concern. This fragment retrieves an API key from the OS keyring (keyctl), parses cloud settings, and (by design/context) likely persists these values into a local .env file. It also clones a remote repository to obtain LFS artifact metadata for a bundled runtime, which is a supply-chain risk if integrity/pinning/verification is not enforced later. No explicit network exfiltration or backdoor execution is shown in the provided snippet, but the credential handling + remote runtime bootstrap are the primary red flags.

This code performs runtime bootstrap and system-level linking: it rewrites shebangs, uses sudo to remove/create symlinks in /usr/local/bin via shell=True subprocess calls, initializes runtime files, and writes an API key and other model settings to an environment file. No explicit exfiltration or reverse shell is present in this snippet, but the combination of privileged shell execution, system path modification, wildcard imports from an unseen module, and persistence of api_key to disk are significant supply-chain/security concerns that warrant deeper review of the `common` module and ENV_FILE permissions.

This fragment is an extraction/bootstrap step that writes to the local filesystem based on an externally supplied archive (`MIRROR_FILE`). The most significant security concern is unsafe ZIP extraction (`zipfile.extractall(BASE_DIR)`) without member path validation, which can enable path traversal/overwrite outside the intended directory if the archive is attacker-controlled. Tar extraction is also performed without additional hardening in this snippet. No overt malware behavior (exfiltration, backdoor, credential theft) is visible here; however, the code creates a high-impact supply-chain/install-stage risk if upstream controls do not ensure the archive’s integrity and safety.

This fragment behaves like a service supervisor/launcher: it detects existing listeners, force-kills processes on specific port ranges, starts a runtime start script, waits for expected ports, and writes discovered port values to an env/config file. It does not directly show classic malware behaviors (no exfiltration/backdoor/reverse shell), but it has high-impact actions (SIGKILL and subprocess execution) and depends on unseen internals from common.py. If common.py or RUNTIME_START were tampered with, it could become a supply-chain/persistence vector.