sglang-prod-incident-triage

Warn

Audited by Socket on Sep 1, 2026

2 alerts found:

Securityx2
SecurityMEDIUM
scripts/replay_trusted_request_dump.py

This module is high-risk primarily due to unsafe use of pickle.load() on .pkl files provided via user-controlled paths. If an attacker can supply or influence those pickle dumps (or if the package processes untrusted dumps), arbitrary code execution is possible during deserialization. The remaining behavior is a straightforward HTTP replay client that forwards dump-derived request content to a user-specified HTTP endpoint and parses/prints response metadata; there is no clear standalone malware/backdoor logic in the snippet itself.

Confidence: 82%Severity: 86%
SecurityMEDIUM
scripts/incident_artifact_tool.py

No direct evidence of covert malware behavior (e.g., reverse shell, persistence, or hidden network exfiltration) is present in this module. However, the tool performs unsafe deserialization: it unpickles arbitrary user-supplied *.pkl files via pickle.load() with no trust boundary enforcement. If dump files are attacker-controlled or not strictly trusted, this can lead to arbitrary code execution at analysis time, making the package materially high risk in that usage mode. The remaining functionality is consistent with diagnostic collection/summarization, with an additional privacy risk from embedding request/prompt previews into locally generated reports.

Confidence: 78%Severity: 80%
Audit Metadata
Analyzed At
Sep 1, 2026, 01:00 AM
Package URL
pkg:socket/skills-sh/bbuf%2Fai-infra-auto-driven-skills%2Fsglang-prod-incident-triage%2F@8fdff4f9bedfe3259073d6ff65c90270407bc527f45885f8967a64a6996b5fe1
Security Audit — socket — sglang-prod-incident-triage