tqx

No clear evidence of embedded malware (no obfuscation/payload, no direct system damage or covert exfiltration to attacker domains). The primary security concern is capability and supply-chain risk: an LLM-driven tool agent that can (1) log in using env credentials, (2) submit arbitrary Python strategy code to a backend via tqx-cli --code, and (3) place/cancel orders and manage workflows by invoking external subprocess commands. Additionally, it sends conversation content/tool schemas to an external LLM endpoint (OpenRouter), increasing privacy and dependency risk. Overall, treat this as high-impact tooling that requires strict policy controls/validation/isolation rather than as confirmed malicious code.

SUSPICIOUS: the skill is mostly aligned with a quant-trading purpose and uses a plausible official CLI path, but it materially increases risk by enabling autonomous live trading, syncing auth tokens into persistent storage, and instructing the agent to fetch additional remote skills from third-party raw URLs. This is high security risk due to real-world financial action capability, not confirmed malware.

No clear evidence of overt malware/backdoor behavior in this module. However, there are notable supply-chain/abuse-relevant security risks: (1) directory traversal/file disclosure via the custom translate_path() implementation, (2) unauthenticated high-impact control endpoints that invoke `tqx-cli` to perform trading/backtesting actions, (3) dynamic reload/execution of an `agent` module based on user input, and (4) leakage of authentication-related metadata (TQX_EMAIL / API-key presence) via /api/status. These issues warrant immediate review/hardening if this service is reachable beyond a trusted local environment.

No direct evidence of overt malware behavior (no networking, subprocess/backdoor patterns, secrets, or exfiltration) is present in this fragment. The security posture is nevertheless weakened by (1) use of an unsanitized run_id directly in filesystem paths for reads/writes, creating potential path traversal/clobbering risk if run_id can be influenced, and (2) runtime import of an unqualified module named agent, which can become a supply-chain/import-resolution execution point. This is best characterized as a local persistence/metrics utility with meaningful abuse and dependency-resolution risks that should be reviewed/mitigated.