@melaya/runner@1.1.29
Run Melaya AI pipelines locally with your own LM Studio or Ollama models
AI Security Review
scanned 2mo ago · by lpm-firewall-aiThis is an explicit but high-risk remote execution agent rather than an install-time malware chain. The code establishes a real service-controlled execution surface and local credential access, but static source does not show automatic installation abuse or unrelated secret exfiltration.
Static reason
One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source
Trigger
The user runs melaya-runner with a runner token and the service sends a runner:run command.
Impact
A compromised or malicious service controlling the runner token can execute arbitrary Python with the user's permissions and access locally available credentials.
Mechanism
Remote code delivery and local Python execution.
AI rationale
This is an explicit but high-risk remote execution agent rather than an install-time malware chain. The code establishes a real service-controlled execution surface and local credential access, but static source does not show automatic installation abuse or unrelated secret exfiltration.
Evidence
dist/sharedVendor.jsdist/connection.jsdist/detect.js~/.melaya-runner/shared/**tmpdir()/melaya-run-${runId}/main.pytmpdir()/melaya-run-${runId}/config.json~/.claude/.credentials.json~/.codex/auth.json~/.config/github-copilot/apps.json~/.config/github-copilot/hosts.json
Network endpoints2
api.melaya.orgapi.github.com
Decision evidence
public snapshotAI called this Suspicious at 91.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for warning
- The runner downloads a server-selected Python bundle and writes its files into the user's Melaya cache.
- A socket command writes server-provided Python into a temporary main.py for execution.
- The bundle and run-code signatures use a key derived from the runner token, so they do not establish an independent code-publisher trust boundary.
- On startup it reads local Claude, Codex, and Copilot credential sources; the Copilot token is sent to GitHub.
Evidence against
- package.json has only a prepublish hook, not an install-time lifecycle hook.
- Execution requires an explicit CLI invocation with a required runner token.
- No source evidence shows secret exfiltration to an unrelated host or automatic execution on package installation.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetworkShellWebSocket
HighEntropyStrings
Source & flagged code
8 flaggeddist/codeWorker.jsView file
51// model (--experimental-permission / --permission): fs read/write
L52: // limited to the scratch dir, child_process and worker_threads denied.
L53: //
High
Child Process
Package source references child process execution.
dist/codeWorker.jsView on unpkg · L51288// ulimit -t: hard CPU time in seconds.
L289: // We use /bin/sh -c "ulimit ...; exec sandbox-exec ..." so both limits
L290: // apply before the Node process starts. This is a soft-then-hard pair;
High
dist/browserProvisioner.jsView file
•matchType = previous_version_dangerous_delta
matchedPackage = @melaya/runner@1.0.93
matchedIdentity = npm:QG1lbGF5YS9ydW5uZXI:1.0.93
similarity = 0.692
summary = stored previous version shares package body but lacks this dangerous source file
Findings
1 Critical6 High4 Medium6 Low
CriticalPrevious Version Dangerous Deltadist/browserProvisioner.js
HighChild Processdist/codeWorker.js
HighShelldist/codeWorker.js
HighSandbox Evasion Gated Capabilitydist/browserProvisioner.js
HighCloud Metadata Accessdist/browserAuthz.js
HighRuntime Package Installdist/lumaLogin.js
HighSemantic Analysis Limiteddist/linkedinLogin.js#virtual:normalized:round1
MediumNetwork
MediumEnvironment Vars
MediumShips Build Helperdist/localRagIngest.py
MediumStructural Risk Force Deep Review
LowNon Install Lifecycle Scripts
LowScripts Present
Affected versions and remediation
This report applies to @melaya/runner@1.1.29.
See all versions of @melaya/runner for other recorded verdicts.
- Review the evidence and your use of @melaya/runner@1.1.29 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.