One DSH model pool for every free quota you have: merge all locally signed-in WorkBuddy accounts with auto-failover, sign in to 13 vendors by QR/web, aggregate 76 vendors' free tiers, plus a usage dashboard, model routing, a local OpenAI-compatible gatewa
Flagged — allowed with a warning
Allowed by default policy, but 17 findings warrant review before installing.
AI security review
scanned · by lpm-firewall-ai
The package's auto-check-in feature is enabled by default and starts with the plugin, then claims credits across configured provider accounts. That matches its advertised automation purpose, but users may not expect account actions to occur automatically, so this warrants a warning; I found no evidence of credential theft or an unrelated collector.
Static reason
One or more suspicious static signals were detected.; source matched previously finalized malicious package; routed for review; previous stored version diff introduced dangerous source
Trigger
Plugin startup schedules auto-check-in; it runs when due unless the saved setting is disabled.
Impact
Daily check-in or reward claims may be made using configured account credentials, creating account activity the user may not expect.
Mechanism
The channel pack obtains providers from the account pool and invokes credits.claimAll for each eligible provider.
Evidence
vendor/channel-pack/pack.js
Network endpoints1
apihub.agnes-ai.com/v1
Decision evidence
public snapshot
Evidence for warning
The channel pack defaults automatic check-in to enabled.
When the plugin starts, it starts auto-check-in, which calls credits.claimAll for providers found in the account pool.
The package describes itself as a multi-vendor quota pool with a points-automation suite.
Evidence against
The postinstall script links host peer dependencies into this package's node_modules; the inspected code shows no external contact from that hook.
Provider API keys are configured alongside their provider endpoints and are added to request headers only for API-key providers.
The package's stated purpose includes account aggregation and points automation.
Behavioral surface
SourceChild process · Crypto · Dynamic require · Environment variables · Filesystem · Network · Web socket
Supply chainHigh entropy strings · URL strings
ManifestNo manifest risk signals triggered.
Scanned 5 files · 5.27 MB of source
50 external domains referenced
127.0.0.1
account.minimax.cn
account.xfinfr.com
accounts.google.com
agent.minimax.cn
agnes-ai.com
ai-gateway.vercel.sh
ai.azure.com
ai.google.dev
aiplatform.googleapis.com
aistudio.google.com
api-overmind.youdao.com
api.anthropic.com
api.cerebras.ai
api.cheaperinference.com
api.cline.bot
api.cloudflare.com
api.cohere.ai
api.deepinfra.com
api.deepseek.com
api.experientiallabs.ai
api.featherless.ai
api.fireworks.ai
api.githubcopilot.com
api.groq.com
api.hunyuan.cloud.tencent.com
api.hyperbolic.xyz
api.inference.wandb.ai
api.kilo.ai
api.kimi.com
api.llm7.io
api.minimax.io
api.minimaxi.com
api.mistral.ai
api.moonshot.ai
api.morphllm.com
api.novita.ai
api.openai.com
api.opper.ai
api.orcarouter.ai
api.perplexity.ai
api.sambanova.ai
api.scaleway.ai
api.siliconflow.com
api.together.ai
api.tokenfactory.nebius.com
api.tokenrouter.com
api.trae.cn
api.trae.com.cn
api.venice.ai
Findings
1 critical · 5 high · 5 medium · 6 low
Where they are, most severe file first. Code opens a file's flagged lines, with their evidence, below.