registry / @proagentstore/cli / 0.4.210
@proagentstore/cli@0.4.210
CLI for creating, publishing, and running ProAgentStore agents
AI Security Review
scanned 18h ago · by lpm-firewall-aiThe flagged execution, network, and persistence behaviors serve the CLI’s agent, browser, and runner functions. The reviewed provider keys stay with their matching provider, and no install-time hook or unsolicited persistence was found.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
User-invoked CLI commands; service installation requires the explicit --install-service option.
Impact
No malicious collection, unrelated credential transfer, or unrequested persistence was identified.
Mechanism
Authenticated requests go to ProAgentStore or the relevant AI provider; local browser and process activity supports requested runner and application workflows.
AI rationale
The flagged execution, network, and persistence behaviors serve the CLI’s agent, browser, and runner functions. The reviewed provider keys stay with their matching provider, and no install-time hook or unsolicited persistence was found.
Network endpoints4
api.proagentstore.onlineapi.anthropic.comapi.openai.comgenerativelanguage.googleapis.com
Decision evidence
public snapshotAI called this Clean at 95.0% confidence as Benign with low false-positive risk.
Evidence for AI clean decision
- The manifest declares no install, preinstall, or postinstall hook; its scripts are build, development, typecheck, and package smoke-test commands.
- Provider API keys are used for model-list requests to their matching provider endpoints, and the code returns model IDs rather than forwarding the keys.
- Local application runs use a run-specific bridge; API-key authentication is refused for that workflow, and named sign-in secrets are scoped to the paused run’s browser session.
- Service persistence is behind the explicit `pags up --install-service` option and requires an existing sign-in.
Evidence against
Behavioral surface
ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystemNetworkShellWebSocket
HighEntropyStringsUrlStrings
Source & flagged code
10 flaggeddist/browser-runner/site-builder/runtime.jsView file
32*/
L33: import { spawn as nodeSpawn } from "node:child_process";
L34: import { randomBytes } from "node:crypto";
High
Child Process
Package source references child process execution.
dist/browser-runner/site-builder/runtime.jsView on unpkg · L32dist/browser-runner/coding/windows-runner-live.jsView file
59busy: async (dir) => {
L60: const ps = spawnSync("powershell", ["-NoProfile", "-Command", `(Get-CimInstance Win32_Process -Filter "Name='Runner.Worker.exe'" | Where-Object { $_.ExecutablePath -like '${dir.rep...
L61: return ps.status === 0 && Number.parseInt(String(ps.stdout).trim(), 10) > 0;
High
Shell
Package source references shell execution.
dist/browser-runner/coding/windows-runner-live.jsView on unpkg · L59Findings
2 Critical5 High5 Medium5 Low
CriticalCommand Output Exfiltrationdist/index.js
CriticalPrevious Version Dangerous Deltadist/browser-runner/local-apply/runtime.js
HighChild Processdist/browser-runner/site-builder/runtime.js
HighShelldist/browser-runner/coding/windows-runner-live.js
HighSame File Env Network Executiondist/index.js
HighSandbox Evasion Gated Capabilitydist/browser-runner/coding/actions-runner.js
HighSemantic Analysis Limiteddist/browser-runner/coding/actions-runner.js#virtual:normalized:round1
MediumDynamic Requiredist/browser-runner/coding/runtime.js
MediumNetwork
MediumEnvironment Vars
MediumInstall Persistencedist/index.js
Affected versions and remediation
This report applies to @proagentstore/cli@0.4.210.
See all versions of @proagentstore/cli for other recorded verdicts.
- Review the evidence and your use of @proagentstore/cli@0.4.210 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.