@mastra/code-sdk@1.10.0
Mastra Code SDK: the agent core behind Mastra Code (everything except the TUI) — build your own UIs and surfaces on top of it
AI Security Review
scanned 1mo ago · by lpm-firewall-aiStatic source inspection establishes conditional forwarding of existing service credentials to an unrestricted configured gateway. The explicit routing gate and absence of concrete malicious behavior support a warning rather than a publication block.
Static reason
One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source
Trigger
Run an Anthropic OAuth model with a gateway-prefixed selection, a configured gateway key, and a custom gateway URL.
Impact
The configured gateway can receive an Anthropic service token and model request content.
Mechanism
The model resolver selects the configured gateway, and its OAuth adapter inserts the stored service token into the authorization header before fetching that URL.
AI rationale
Static source inspection establishes conditional forwarding of existing service credentials to an unrestricted configured gateway. The explicit routing gate and absence of concrete malicious behavior support a warning rather than a publication block.
Evidence
dist/index.jsdist/agents/model.jsdist/agents/mastracode-gateway.jsdist/providers/claude-max.jsdist/auth/storage.jsauth.json
Network endpoints1
gateway-api.mastra.ai
Decision evidence
public snapshotAI called this Suspicious at 96.0% confidence as Dangerous Capability with medium false-positive risk.
Evidence for warning
- The active agent entrypoint delegates model selection to the dynamic model resolver.
- The resolver accepts a gateway URL from settings or MASTRA_GATEWAY_URL without a destination allowlist.
- Gateway routing requires both a gateway key and a gateway-prefixed model selection.
- The gateway sends Anthropic OAuth models to its configured URL using the OAuth fetch adapter.
- The adapter retrieves the stored Anthropic token and sends it as a bearer credential to the supplied URL.
- Credentials are loaded from the application's auth.json; package.json has no registry install hooks.
Evidence against
- Gateway forwarding is conditional on model selection and gateway authentication.
- No covert recipient, malicious install behavior, or malware intent was established.
Behavioral surface
ChildProcessCryptoDynamicRequireEnvironmentVarsEvalFilesystemNetworkShell
Source & flagged code
4 flaggeddist/plugins/loader.jsView file
117url.searchParams.set("mtimeNs", stat.mtimeNs.toString());
L118: const mod = await import(url.href);
L119: return validatePluginExport(mod.default ?? mod.plugin);
Medium
Dynamic Require
Package source references dynamic require/import behavior.
dist/plugins/loader.jsView on unpkg · L117dist/utils/storage-maintenance.jsView file
1import { getDatabasePath, getVectorDatabasePath } from "./project.js";
L2: import { dirname } from "path";
Medium
Unsafe Vm Context
Package source executes code through a VM context API.
dist/utils/storage-maintenance.jsView on unpkg · L1dist/index.jsView file
Findings
2 High5 Medium6 Low
HighSandbox Evasion Gated Capabilitydist/index.js
HighPrevious Version Dangerous Deltadist/utils/project.js
MediumDynamic Requiredist/plugins/loader.js
MediumUnsafe Vm Contextdist/utils/storage-maintenance.js
MediumNetwork
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
LowScripts Present
LowEval
LowFilesystem
LowHigh Entropy Strings
LowTelemetry
LowUrl Strings
Affected versions and remediation
This report applies to @mastra/code-sdk@1.10.0.
See all versions of @mastra/code-sdk for other recorded verdicts.
- Review the evidence and your use of @mastra/code-sdk@1.10.0 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.