luzra-cli@0.1.12
터미널 AI(Claude Code) 프롬프트 습관을 채점·코칭하는 리포트 — 세션 로그는 로컬에서만 분석됩니다.
AI Security Review
scanned 3mo ago · by lpm-firewall-aiSource establishes an explicit-user scheduled, self-updating analytics agent with remotely supplied instructions, but not malicious install-time mutation, exfiltration of raw prompts, or covert payload execution. Warn rather than block.
Static reason
One or more suspicious static signals were detected.
Trigger
User runs `luzra init`; scheduled reports subsequently run daily.
Impact
A compromised dashboard or later npm release could influence recurring execution; derived session analytics are uploaded after opt-in connection.
Mechanism
User-authorized scheduling, self-update, Claude CLI invocation, and dashboard synchronization.
AI rationale
Source establishes an explicit-user scheduled, self-updating analytics agent with remotely supplied instructions, but not malicious install-time mutation, exfiltration of raw prompts, or covert payload execution. Warn rather than block.
Evidence
package.jsonbin/luzra.jssrc/init.jssrc/install.jssrc/schedule.jssrc/engine.jssrc/sync.jssrc/masking.js~/.claude/projects~/.luzra/config.json~/.luzra/reports/profile.json~/Library/LaunchAgents/ai.luzra.daily.plist~/.config/systemd/user/luzra.service~/.config/systemd/user/luzra.timer
Network endpoints2
luzra.aimgonna.comregistry.npmjs.org/luzra-cli/latest
Decision evidence
public snapshotAI called this Suspicious at 90.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for warning
- src/init.js installs a daily OS scheduler after interactive init.
- src/install.js self-updates global luzra-cli from npm during reports.
- src/engine.js sends server-supplied judge instructions to local `claude -p`.
- src/sync.js uploads score/profile data after dashboard connection.
Evidence against
- package.json has no lifecycle scripts.
- All persistence is reached from explicit `luzra init`, not installation.
- Session prompts are masked before being passed to the judge.
- No hidden payload, eval, credential harvesting, or foreign agent-config writes found.
Behavioral surface
ChildProcessEnvironmentVarsFilesystemNetworkShell
HighEntropyStringsUrlStrings
NoLicense
Source & flagged code
6 flaggedsrc/connect.jsView file
1// 디바이스 코드 페어링 — 토큰을 사람이 복사·전달할 필요가 없다.
L2: import { spawn } from "node:child_process";
L3: import { c, blank, info, ok, warn } from "./ui.js";
High
1// 디바이스 코드 페어링 — 토큰을 사람이 복사·전달할 필요가 없다.
L2: import { spawn } from "node:child_process";
L3: import { c, blank, info, ok, warn } from "./ui.js";
L4:
L5: export const DEFAULT_URL = "https://luzra.aimgonna.com";
L6:
...
L8: const cmd =
L9: process.platform === "win32" ? ["cmd", ["/c", "start", "", url]]
L10: : process.platform === "darwin" ? ["open", [url]]
...
L23: headers: { "Content-Type": "application/json" },
L24: body: JSON.stringify(body || {}),
L25: });
High
Sandbox Evasion Gated Capability
Source gates dangerous network, credential, or execution behavior behind CI, host, platform, time, or geo fingerprint checks.
src/connect.jsView on unpkg · L1src/install.jsView file
17// Windows의 npm은 npm.cmd라 shell이 필요하다. 인자를 배열로 주면 경고가 나므로 문자열로 넘긴다.
L18: const r = spawnSync("npm root -g", { encoding: "utf-8", shell: true });
L19: const dir = (r.stdout || "").trim();
Findings
5 High4 Medium4 Low
HighChild Processsrc/connect.js
HighShellsrc/install.js
HighSandbox Evasion Gated Capabilitysrc/connect.js
HighTrigger Reachable Persistencesrc/schedule.js
HighRuntime Package Installsrc/install.js
MediumNetwork
MediumEnvironment Vars
MediumInstall Persistencesrc/schedule.js
MediumStructural Risk Force Deep Review
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
LowNo License
Affected versions and remediation
This report applies to luzra-cli@0.1.12.
See all versions of luzra-cli for other recorded verdicts.
- Review the evidence and your use of luzra-cli@0.1.12 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.