@wayai/cli@0.3.153
WayAI CLI — sync hub configuration between local files and the platform
AI Security Review
scanned 27d ago · by lpm-firewall-aiThe source supports a warning for an explicit command that modifies agent skill configuration, while showing no npm install hook or confirmed attack behavior. Classify this as an agent capability risk, not a block-worthy attack.
Static reason
One or more suspicious static signals were detected.
Trigger
A user invokes the CLI skill-install command.
Impact
The installed skill becomes available to the selected agent after reload; the inspected code does not establish malicious behavior in the returned files.
Mechanism
The command authenticates, retrieves skill files from the WayAI API, and writes them under the project’s detected agent skill directories.
AI rationale
The source supports a warning for an explicit command that modifies agent skill configuration, while showing no npm install hook or confirmed attack behavior. Classify this as an agent capability risk, not a block-worthy attack.
Evidence
dist/index.js.claude/skills/<skill>/SKILL.md.opencode/skills/<skill>/SKILL.md.agents/skills/<skill>/SKILL.md
Network endpoints1
api.wayai.pro
Decision evidence
public snapshotAI called this Suspicious at 86.0% confidence as Dangerous Capability with medium false-positive risk.
Evidence for warning
- The manifest defines no install lifecycle hook; its scripts are build, development, typecheck, and test commands.
- The bundle identifies WayAI skill files in Claude, OpenCode, and agent skill directories.
- The explicit skill-install command requires authentication, fetches skill files from the WayAI API, and writes them into supported agent skill directories.
Evidence against
- The package exposes a CLI entrypoint, and the skill installation runs through an explicit command rather than an npm lifecycle hook.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell
HighEntropyStringsUrlStrings
Source & flagged code
4 flaggeddist/index.jsView file
14636});
L14637: import { execFileSync } from "child_process";
L14638: import * as fs2 from "fs";
High
51while (cursor < text.length) {
L52: while (cursor < text.length && !isBase64UrlChar(text[cursor])) cursor++;
L53: const headerStart = cursor;
...
L236: "use strict";
L237: CREDENTIAL_KEY_RE = /token|secret|credential|password|passwd|api[_-]?key|authorization|private[_-]?key|session[_-]?id|cookie|jwt/i;
L238: NUMERIC_CREDENTIAL_KEY_RE = /(?:password|passwd|pin|token|secret|credential|api[_-]?key|authorization|private[_-]?key|session[_-]?id|cookie|jwt)$/i;
...
L249: [/\bway_[A-Za-z0-9_-]{8,}/g, "way_[REDACTED]"],
L250: // wst_ single-use WebSocket tickets (same opaque base64url format).
L251: [/\bwst_[A-Za-z0-9_-]{8,}/g, "wst_[REDACTED]"],
...
L366: try {
L367: const __dirname2 = dirname(fileURLToPath(import.meta.url));
L368: const pkg2 = JSON.parse(readFileSync(join(__dirname2, "..", "
High
Sandbox Evasion Gated Capability
Source gates dangerous network, credential, or execution behavior behind CI, host, platform, time, or geo fingerprint checks.
dist/index.jsView on unpkg · L5116018package = @wayai/cli; repositoryIdentity = platform; dependency = @napi-rs/keyring
L16018: try {
L16019: const mod = await import("@napi-rs/keyring");
L16020: const Entry = mod.Entry;
Findings
5 High3 Medium4 Low
HighChild Processdist/index.js
HighShell
HighSandbox Evasion Gated Capabilitydist/index.js
HighCopied Package Dependency Bridgedist/index.js
HighRuntime Package Installdist/index.js
MediumNetwork
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
Affected versions and remediation
This report applies to @wayai/cli@0.3.153.
See all versions of @wayai/cli for other recorded verdicts.
- Review the evidence and your use of @wayai/cli@0.3.153 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.