magpie-trustoken-auth@0.3.6
Trustoken / AiPy for magpie — phone-code sign-in, the account's live model list, 灵石 balance, auto check-in and daily-task rewards. DeepSeek / Qwen / GLM / Doubao / Kimi / Hunyuan.
AI Security Review
scanned 5d ago · by lpm-firewall-aiThe package automatically creates and runs model tasks during check-in, which can consume the user's quota without per-task consent. The tasks are tied to the advertised reward flow and use a local AiPy endpoint, but the automatic resource spending meets the malicious-action rule.
Decision evidence
public snapshot- The package enables magpie auto-check-in by default.
- The check-in hook checks for incomplete daily tasks and automatically calls ensureAipyTasks to create AiPy sessions.
- Each created session sends force:true and a model to AiPy's local task API; the hook chooses the cheapest model, but still triggers model work that can consume the user's quota.
- The hook is wired into the plugin and obtains the current account credentials before performing check-in actions.
- The task creation is limited to reported gaps in Trustoken daily tasks.
- The request is sent to a loopback AiPy endpoint using the local AiPy API key, rather than sending that key to a remote host.
- The package presents itself as a Trustoken/AiPy plugin for check-in and daily-task rewards.
Source & flagged code
4 flaggedPackage source references dynamic require/import behavior.
scripts/check.mjsView on unpkg · L75Findings
Affected versions and remediation
This report applies to magpie-trustoken-auth@0.3.6.
See all versions of magpie-trustoken-auth for other recorded verdicts.
- Avoid installing magpie-trustoken-auth@0.3.6. Remove it from direct dependencies and check your lockfile for transitive copies.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
- If this version ran, investigate the affected machine and build environment. Rotate credentials it could access and rebuild from a trusted environment.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.