Back to skill

Security audit

wukong-sec-canary-7f3a

Security checks for vulnerabilities and agentic risk

Overview

The skill claims to be a no-op canary responder, but the package also contains unrelated web app files and install/publish instructions that are not explained.

Treat this as a Review item, not a simple canary. Install only if the publisher removes or clearly explains the extra web app files, and do not run the displayed global npm, login, or publish commands unless you independently trust and pin that tooling.

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 (1)

T08 · Insecure Dependencies

Warning
Location
index.html:56
Finding

Unpinned Global npm Package Installation Instruction

Content
View full analysis
$ npm i -g clawhub ``` ### Technical Analysis The bundled page presents a copyable command that installs the latest available `clawhub` npm package globally. The dependency is not pinned to an audited version and has no integrity constraint. npm installation can execute package lifecycle scripts with the invoking user's privileges. Because the package version is mutable, the code ultimately installed may differ from the code available when this Skill was reviewed. This creates a supply-chain risk if the package, its dependencies, its publisher account, or the package registry is compromised. The instruction is also unrelated to the functionality declared in `SKILL.md`, which describes a side-effect-free fixed-token responder requiring no network access, package installation, filesystem modification, or command execution. It therefore exceeds the minimum capabilities necessary for the declared Skill. The command is displayed as documentation and is not automatically executed by the submitted code. Exploitation requires a user or automation system to copy and run it. ### Attack Path 1. A user opens or otherwise trusts the bundled `index.html` page. 2. The page presents the copyable `npm i -g clawhub` command. 3. An attacker publishes a compromised release, compromises the publisher or registry, or introduces a malicious transitive dependency. 4. The user executes the unpinned command. 5. npm retrieves the then-current package and dependencies rather than a previously audited version. 6. Malicious package or lifecycle-script code executes with the permissions of the user running npm. 7. The package may modify user files, global Node.js tooling, shell- ...[truncated 851 chars]
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (7)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared description says this skill should act as a deterministic status responder for integration smoke tests, implying a simple function that returns a fixed token and little else. The supplied code instead contains minified frontend utility code: React JSX/runtime helpers, SVG icon component construction, context usage, class-name concatenation, camel/kebab-case transformations, accessibility checks, and a proxy that builds component reference paths. There is no visible logic for handling heartbeat requests, returning a fixed status string, or acting as a smoke-test responder. This is a materially different primary purpose, so the description does not accurately represent the code.

Content

No source excerpt is available for this finding.

Hidden Instructions

High
Category
Prompt Injection
Confidence
90% confidence
Finding

The file contains large volumes of visually noisy, partially hidden or non-primary content such as repeated ASCII blocks and framework serialization that can bury meaningful behavior and impede review. In the context of a supposed fixed-string responder, this constitutes concealment-by-complexity and increases the likelihood that risky functionality is overlooked during manual or automated inspection.

Content

Scanner excerpt · index.html (reported line 1)May include surrounding context.

html
<!DOCTYPE html><html data-theme-family="claw" data-theme-mode="system" lang="en"><head><meta charSet="utf-8" nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta name="viewport" content="width=device-width, initial-scale=1" nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><link rel="preload" as="image" href="/logo-transparent.png"/><link rel="stylesheet" href="/assets/styles-BP2P2THE.css" nonce="bXerb/4uKHtpgBNwcGwmvDzX" data-precedence="default"/><link rel="stylesheet" href="/assets/design-system-BQ18hLKc.css" nonce="bXerb/4uKHtpgBNwcGwmvDzX" data-precedence="default"/><meta property="csp-nonce" content="bXerb/4uKHtpgBNwcGwmvDzX"/><title>ClawHub</title><meta name="description" content="ClawHub — a fast skill registry for agents, with vector search." nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta property="og:site_name" content="ClawHub" nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta property="og:type" content="website" nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta property="og:title" content="ClawHub" nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta property="og:description" content="ClawHub — a fast skill registry for agents, with vector search." nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta property="og:image" content="https://clawhub.ai/og.png?v=20260723-1" nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta property="og:image:width" content="1200" nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta property="og:image:height" content="630" nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta property="og:image:alt" content="ClawHub — ClawHub — a fast skill registry for agents, with vector search." nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta name="twitter:card" content="summary_large_image" nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta name="twitter:title" content="ClawHub" nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta name="twitter:description" content="ClawHub — a fast skill registry for agents, with vector search." nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta name="twitter:image" content="https://clawhub.ai/og.png?v=20260723-1" nonce="bXerb/4uKHtpgBNwcGwmvDzX"/><meta name="twitt
...[truncated 27 chars]

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The declared skill is a fixed-string heartbeat responder, but the file is a full marketplace web application with catalog browsing, publishing flows, imports, external links, hydration state, and client-side runtime loading. This severe capability/intent mismatch indicates deceptive packaging and materially enlarges the attack surface beyond what operators would reasonably authorize for a smoke-test canary skill.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The page exposes unjustified catalog, publishing, import, and ecosystem-browsing capabilities that are unnecessary for a smoke-test heartbeat responder. Even if some actions are only UI elements in this file, bundling these pathways encourages unauthorized user interaction, loads additional code, and creates opportunities for abuse, phishing, data exposure, or later activation through hydrated client modules.

Content

No source excerpt is available for this finding.

Hidden Instructions

High
Category
Prompt Injection
Confidence
89% confidence
Finding

The second hidden-instructions finding reflects additional concealed or low-visibility bulk content around ecosystem, publish, and import messaging embedded in decorative/footer regions. While not executable by itself, this kind of buried content is suspicious in a canary skill and can be used to distract reviewers, carry covert instructions, or normalize unrelated high-risk capabilities.

Content

Scanner excerpt · index.html (reported line 56)May include surrounding context.

html
hooks runners slash-commands skill.md templates scanners review-bots     safe browse paths   official gateways   publisher handles   org trust   &gt;&gt;&gt; install scan publish verify    @@ gateway @@ registry @@ agents @@   openclaw ecosystem    crabbox clickclack crawler packs gateway plugins     30 skills 12 plugins    /api/v1/skills   /owners   /audit   /ship
openclaw ecosystem    crabbox clickclack crawler packs gateway plugins   ....:: clawhub/openclaw ::....  skills plugins publishers trust signals     30 skills 12 plugins    /api/v1/skills   /owners   /audit   /ship   ---- downloads installs stars lineage ownership docs package integrity   :::: signed manifests ::::: moderated releases ::::: version history ::::
---- downloads installs stars lineage ownership docs package integrity   &gt;&gt;&gt; install scan publish verify    @@ gateway @@ registry @@ agents @@   :::: signed manifests ::::: moderated releases ::::: version history ::::     safe browse paths   official gateways   publisher handles   org trust     hooks runners slash-commands skill.md templates scanners review-bots
  safe browse paths   official gateways   publisher handles   org trust     30 skills 12 plugins    /api/v1/skills   /owners   /audit   /ship     hooks runners slash-commands skill.md templates scanners review-bots   ....:: clawhub/openclaw ::....  skills plugins publishers trust signals   openclaw ecosystem    crabbox clickclack crawler packs gateway plugins</pre><div class="home-v2-byos-reveal-image home-v2-byos-reveal-image--top"></div></div><div class="home-v2-byos-content"><header class="home-v2-byos-head oc-section-heading"><span class="home-v2-byos-eyebrow oc-eyebrow">ClawHub CLI</span><h2 id="home-v2-byos-title" class="home-v2-byos-title oc-section-title">Bring your skills to ClawHub</h2><p class="home-v2-byos-lede oc-section-copy">Publish and sync your skills to ClawHub, your way.</p></header><div class="home-v2-byos-tabs" role="tablist" aria-label="Choose an audience"><
...[truncated 27 chars]

Ae3

High
Category
analysis-evasion
Confidence
94% confidence
Finding

The embedded NUL bytes appear inside the serialized router state identifiers in the inline script, which is unusual for a static HTML skill artifact and can confuse downstream parsers, scanners, log pipelines, or file-processing logic. In a skill distribution context, malformed text artifacts increase the risk of analysis evasion and inconsistent behavior across tooling, especially when the skill already deviates substantially from its declared purpose.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The root HTML element sets lang="en", and the page content is presented in English as the default experience with no visible language selector or opt-in. Under the stated policy, hard-coding a specific language without offering a user choice is a natural-language locale policy issue.

Content

No source excerpt is available for this finding.