@sokeai/cli@1.0.104
授客AI官方CLI工具 - 支持AI Agent Skills
AI Security Review
scanned 24d ago · by lpm-firewall-aiThis is a remote staged-payload design activated by installation and later command use, so the absent executable prevents a clean source-only assessment. The inspected scripts do not show credential theft, destructive actions, or active agent-control mutation.
Static reason
High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review
Trigger
npm postinstall downloads the binary; a later soke-cli invocation starts it.
Impact
A compromised selected binary source could supply code executed when the CLI is used; static review cannot inspect that payload.
Mechanism
The installer selects a remote source, writes the downloaded binary under bin, and the launcher spawns it.
AI rationale
This is a remote staged-payload design activated by installation and later command use, so the absent executable prevents a clean source-only assessment. The inspected scripts do not show credential theft, destructive actions, or active agent-control mutation.
Evidence
scripts/install.jsscripts/run.jsbin/soke-clibin/soke-cli.exe
Network endpoints2
soke-frontendstatic-cdn.soke.cn/soke-cli/releasesgithub.com/sokeai/soke-cli/releases/download
Decision evidence
public snapshotAI called this Suspicious at 90.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for warning
- The manifest runs scripts/install.js automatically after installation and maps the command to scripts/run.js.
- The installer builds download sources from an environment-configured host and a package default.
- The installer downloads a platform binary into its bin directory during postinstall.
- The command launcher starts that downloaded binary with the caller's arguments and environment.
Evidence against
- The installer verifies a SHA-256 value obtained from each selected source before moving the binary into place.
- The agent-workspace synchronization and GUI setup code is commented out and does not run.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell
HighEntropyStringsUrlStrings
Source & flagged code
7 flagged · loading sourcepackage.jsonView file
•scripts.postinstall = node scripts/install.js
High
Install Time Lifecycle Scripts
Package defines install-time lifecycle scripts.
package.jsonView on unpkg•scripts.postinstall = node scripts/install.js
Medium
Ambiguous Install Lifecycle Script
Install-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgscripts/install.jsView file
14if (!fs.existsSync(srcDir)) return;
L15: if (!fs.existsSync(destDir)) fs.mkdirSync(destDir, { recursive: true });
L16:
...
L37:
L38: fs.copyFileSync(srcPath, destPath);
L39: }
...
L72: const config = JSON.parse(configText);
L73: const workspaceDir = config?.defaults?.agents?.openaiAgents?.main?.workspace;
L74: if (typeof workspaceDir === 'string' && workspaceDir.length > 0) {
...
L120: try {
L121: fs.mkdirSync(targetDir, { recursive: true });
L122: for (const skillName of skillNames) {
Critical
Findings
1 Critical4 High5 Medium4 Low
CriticalAi Agent Control Hijackscripts/install.js
HighInstall Time Lifecycle Scriptspackage.json
HighInstall Named Payload Filescripts/install.js
HighSemantic Analysis Limitedscripts/run.js
HighKnown Malware Source Similarityscripts/run.js
MediumAmbiguous Install Lifecycle Scriptpackage.json
MediumNetwork
MediumEnvironment Vars
MediumShips Build Helperskills/multi/soke-ai-coach-director/scripts/gate_check.py
MediumStructural Risk Force Deep Review
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
Low
Affected versions and remediation
This report applies to @sokeai/cli@1.0.104.
See all versions of @sokeai/cli for other recorded verdicts.
- Review the evidence and your use of @sokeai/cli@1.0.104 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.