Back to skill

Security audit

AI 内容检测 AI Detect

Security checks across malware telemetry and agentic risk

Overview

This appears to be a legitimate AI-media detection skill, but it needs Review because it can upload local media to dLazy services and relies on an externally installed npm CLI.

Install only if you are comfortable sending selected media to dLazy/Hive-backed services and storing a dLazy API key locally. Prefer reviewing the @dlazy/cli package/source first, avoid global installation when possible, and confirm each local file before analysis, especially for private, biometric, client, or regulated media.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:13
Finding
Unreviewed Third-Party npm CLI Installation and Execution<![CDATA[ ## Vulnerability Details **File Location**: `SKILL.md`, lines 13–14 **Vulnerability Type**: Third-party supply-chain exposure through an unaudited npm package **Risk Level**: Medium ### Vulnerable Code ```yaml 'install': 'npm install -g @dlazy/cli@1.2.3', 'installAlternative': 'npx @dlazy/cli@1.2.3', ``` ### Technical Analysis The Skill directs the environment to download and execute the third-party package `@dlazy/cli` from the npm registry. Although the package version is pinned to `1.2.3`, the audited project does not contain the package's source code, an npm lockfile, a verified integrity hash, a signed provenance record, or a vendored copy that would allow its effective behavior to be reviewed. The global installation option can execute npm lifecycle scripts during installation and leaves the CLI available in the user's environment after the Skill run. The `npx` alternative also downloads and runs code that is outside the reviewed project boundary. This is a supply-chain trust issue rather than evidence that the named package is currently malicious. Exploitation depends on compromise of the package, its publisher account, the registry delivery path, or another dependency included by the package. ### Attack Path 1. An attacker compromises the npm package, its publisher account, or a transitive dependency delivered with the referenced release. 2. A user or agent loads the Skill and follows its installation metadata. 3. The environment runs either `npm install -g @dlazy/cli@1.2.3` or `npx @dlazy/cli@1.2.3`. 4. npm retrieves code that was not included in this audit. 5. Malicious lifecycle or runtime code executes with the permissions of the invoking user. 6. The code could access files and environment variables available to that user, including media supplied to the CLI and potentially the configured dLazy API credential. ### Impact Assessment Successful exploitation could provide arbitrary code execution with the privileges of the user runn ...[truncated 627 chars]
Remediation
<![CDATA[ ## Remediation Suggestions 1. Vendor the exact CLI source or executable artifact into a reviewable, version-controlled release process. 2. Commit an npm lockfile and verify package integrity using the registry-provided SHA-512 integrity value. 3. Require signed package provenance and verify the publisher and release workflow before installation. 4. Prefer a project-local installation over `npm install -g` to limit persistent changes to the user's environment. 5. Disable npm lifecycle scripts where they are unnecessary, for example by using `npm install --ignore-scripts` after confirming that the package does not legitimately require them. 6. Execute the CLI in a sandbox or container with: - Read-only access to unrelated files. - Access only to explicitly selected media. - A restricted environment-variable set. - Network access limited to documented dLazy endpoints. - No administrative or root privileges. 7. Document and verify the package's transitive dependencies as part of the Skill release process. 8. Add automated dependency scanning and alerting for ownership changes, provenance failures, known vulnerabilities, and unexpected package-content changes. ]]>

T08 · Insecure Dependencies

Warning
Location
SKILL-cn.md:13
Finding
Unreviewed Third-Party npm CLI Installation and Execution in Chinese Skill Definition<![CDATA[ ## Vulnerability Details **File Location**: `SKILL-cn.md`, lines 13–14 **Vulnerability Type**: Third-party supply-chain exposure through an unaudited npm package **Risk Level**: Medium ### Vulnerable Code ```yaml 'install': 'npm install -g @dlazy/cli@1.2.3', 'installAlternative': 'npx @dlazy/cli@1.2.3', ``` ### Technical Analysis The localized Skill definition contains the same instructions to retrieve and execute `@dlazy/cli@1.2.3` from npm. Version pinning reduces unintended upgrades but does not independently establish the integrity or safety of the downloaded artifact. The project contains no source implementation, lockfile, integrity hash, or verifiable provenance record for the external CLI. Both installation methods cross the audited trust boundary. The global installation method can run package lifecycle scripts and persist the installed command in the user's environment, while the `npx` method downloads and immediately executes external package code. This finding identifies supply-chain exposure and does not assert that the referenced package is currently malicious. ### Attack Path 1. An attacker compromises the referenced npm release, its publisher infrastructure, or one of its packaged dependencies. 2. A user or agent follows the installation metadata in `SKILL-cn.md`. 3. npm downloads the external package. 4. Installation lifecycle code or the subsequently invoked CLI runs with the current user's permissions. 5. Compromised code accesses user-readable resources or secrets and may transmit them over the network. ### Impact Assessment A successful package compromise could result in arbitrary user-level code execution. Potentially exposed resources include local files, environment variables, uploaded media, and the dLazy API key stored in the user's configuration. A global installation can additionally alter the user's persistent command environment. The reviewed project provides no evidence of current malicious behavior in the depe ...[truncated 100 chars]
Remediation
<![CDATA[ ## Remediation Suggestions 1. Apply the same dependency controls to every localized Skill definition so users receive identical security guarantees. 2. Replace global installation guidance with a project-local, locked installation. 3. Pin and verify the package using a lockfile and SHA-512 integrity metadata. 4. Verify signed npm provenance and monitor publisher ownership and release-workflow changes. 5. Vendor or otherwise make the exact executable implementation available for audit. 6. Run the dependency without elevated privileges and inside a restricted sandbox. 7. Restrict filesystem access to explicitly selected media and restrict network access to documented service endpoints. 8. Avoid exposing unrelated environment variables or credentials to the CLI process. 9. Continuously scan direct and transitive dependencies for known vulnerabilities and unexpected artifact changes. ]]>

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Vague Triggers

Medium
Confidence
85% confidence
Finding
The trigger list includes very broad terms like "detect" and "deepfake" that can match ordinary user conversation and cause unintended invocation of this skill. Because this skill can upload local media to external services and initiate paid API calls, accidental activation increases privacy and cost risk beyond a harmless routing error.

Missing User Warnings

High
Confidence
97% confidence
Finding
The skill states that local paths may be passed directly and that the CLI will automatically upload files to dLazy media storage and then send them to backend services, but it does not present this as a prominent user warning before use. This creates a real privacy and data-handling risk because users or agents may submit sensitive local images, videos, or audio without clear, informed consent that the content leaves the local environment and is stored/processed by third parties.

Vague Triggers

Medium
Confidence
93% confidence
Finding
The trigger list includes very generic terms such as 'detect' and 'is this AI', which can match ordinary conversation and cause the agent to invoke this skill unexpectedly. In this skill's context, accidental invocation is more dangerous because the tool may upload local media or submit user-provided URLs to a remote service, creating privacy, cost, and data-handling risks without sufficiently explicit user intent.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The skill explains in the implementation details that local files are uploaded to dLazy's remote storage and then analyzed by a hosted API, but it does not present this as a prominent user warning before use. Because this tool handles images, videos, and audio—potentially highly sensitive biometric or personal data—users may unknowingly cause exfiltration of local media to third-party infrastructure.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.