soia-pkm-alipan-curator

The skill's documented behaviors mostly match its stated Aliyun Drive curation purpose, and it emphasizes confirmations, audit trails, and avoiding secret leakage. The main concerns are transitive trust in the hard dependency, the skill-to-skill installation pattern, and the wildcard permission request; these make the skill suspicious from an agent-security perspective even without clear malicious intent.

No direct malicious payload indicators (e.g., exfiltration, reverse shells, cryptomining) are evident in this fragment. The primary security concern is high-impact orchestration dependent on dynamic execution/trust boundaries: the companion module is executed at import time via exec_module, and the external runner path can be overridden via SOIA_ALIPAN_RUNNER, enabling arbitrary code execution if an attacker influences the environment or package contents. Overall, treat as a powerful migration controller that is safe only when the packaged companion module and the resolved runner are fully trusted and protected; otherwise the security risk is moderate to high due to potential execution hijack.

No direct malicious behavior (e.g., exfiltration, keylogging, embedded payloads) is evident in this Python fragment. The main security risk is supply-chain/execution surface: it runs a Node.js builder script via subprocess with environment-controlled NODE_OPTIONS and user-influenced node/artifact-runtime paths, and then overwrites/deletes Excel outputs based on a plan. This warrants review of build_workbooks.mjs and the artifact-runtime dependency tree for any malicious or unexpected actions.