Back to skill

Security audit

SEO Expert

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed SEO reference skill with optional HTTP checks and no executable runtime component; the main cautions are example telemetry code and an unpinned maintainer packaging command.

Install this if you want a German Google Search Central based SEO reference skill. Allow `curl` only when you want the agent to perform user-directed HTTP checks against a site. Do not paste real API keys or tokens into chat, and review any copied JavaScript logging examples for privacy before deploying them. Maintainers should pin `skills-ref` or use a local locked validator before packaging.

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
README.md:18
Finding

Unpinned npm Package Execution in the Documented Packaging Workflow

Content
View full analysis

Vulnerability Details

File Location: README.md:18-22
Vulnerability Type: Unpinned third-party package execution
Risk Level: Medium

Vulnerable Code

bash
./scripts/package-seo-expert-for-clawhub.sh

Output: build/clawhub-publish/seo-expert-pro. Runs npx skills-ref validate.

Technical Analysis

The documented packaging workflow runs npx skills-ref validate without specifying an audited package version or integrity constraint. If skills-ref is unavailable locally, npx may download the current package version and execute its code. The code executed by this workflow can therefore change after the Skill has been reviewed.

This creates a supply-chain risk because control over the resolved npm package or its dependencies could allow arbitrary code to run under the maintainer account. The referenced wrapper script, scripts/package-seo-expert-for-clawhub.sh, is not included in the audited artifact, so any version pinning, registry restrictions, or other safeguards inside it could not be verified.

Attack Path

  1. An attacker compromises the skills-ref package, one of its transitive dependencies, or the registry resolution path.
  2. The attacker publishes a malicious version that is eligible for unpinned resolution.
  3. A maintainer follows the README and runs ./scripts/package-seo-expert-for-clawhub.sh.
  4. The wrapper invokes npx skills-ref validate.
  5. If no trusted local installation is selected, npx retrieves and executes the attacker-controlled package.
  6. The malicious package runs with the invoking maintainer's operating-system privileges and can access resources available to that process.

Impact Assessment

Successful exploitation could execute arbitrary code with the privileges of the maintainer running the packaging workflow. Depending on that account's environment and sandboxing, this may permit:

  • Reading or modifying files accessible to the maintain ...[truncated 499 chars]
Remediation
View remediation

Remediation Suggestions

  1. Add skills-ref as a pinned development dependency, using an exact audited version rather than a version range.
  2. Commit the package lockfile and use a deterministic clean installation command such as npm ci.
  3. Invoke only the installed local binary, for example:
    bash
    npx --no-install skills-ref validate
    
    Alternatively, call the binary directly from node_modules/.bin.
  4. Enable lockfile integrity verification and use a trusted, explicitly configured npm registry.
  5. Review and update the package and its transitive dependencies through a controlled dependency-update process.
  6. Run packaging in a restricted CI environment without production credentials, unnecessary environment variables, or write access outside the build workspace.
  7. Include scripts/package-seo-expert-for-clawhub.sh in the auditable source distribution so its exact behavior and safeguards can be reviewed.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (39)

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · references/02_crawling_indexierung.md (reported line 173)May include surrounding context.

Das Meta-Tag wird im <head>-Bereich der HTML-Seite platziert:

html
<!-- Seite nicht indexieren -->
<meta name="robots" content="noindex">

<!-- Seite nicht indexieren und Links nicht folgen -->

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · references/02_crawling_indexierung.md (reported line 559)May include surrounding context.

```

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · references/04_monitoring_fehlerbehebung.md (reported line 152)May include surrounding context.

Lösung:

html
<!-- Falsch: -->
<div>Seite nicht gefunden</div>
<!-- Liefert HTTP 200 -->

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · references/04_monitoring_fehlerbehebung.md (reported line 609)May include surrounding context.

md
<!-- Bilder optimieren -->
<img src="image.webp" width="800" height="600" alt="...">

<!-- Preload für kritische Ressourcen -->
<link rel="preload" as="image" href="hero-image.webp">

<!-- Critical CSS inline -->

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · references/04_monitoring_fehlerbehebung.md (reported line 812)May include surrounding context.

md
<!-- Viewport Meta-Tag -->
<meta name="viewport" content="width=device-width, initial-scale=1">

<!-- Responsive Images -->
<img src="image.jpg" style="max-width: 100%; height: auto;">

<!-- Touch-Targets -->

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The file specifies that the references are German Google Search Central excerpts, but it does not offer a language choice or explain that German is required for a region-specific use case. This creates a language/locale constraint in the skill's natural-language instructions without explicit user opt-in.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The instruction says to use the skill for 'SEO-related work' and includes broad examples like 'content quality for search' and 'clarifying what is vs is not in scope.' This lacks clear trigger boundaries or negative examples, making unintended invocation for general content or analytics tasks more likely.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The manifest describes an SEO expertise skill focused on technical SEO, content SEO, site structure, and measurement. This file goes far beyond general SEO guidance and embeds large, domain-specific implementation guidance for job postings, Q&A, local business, software applications, events, recipes, and courses, which broadens the skill into many unrelated content verticals rather than staying narrowly aligned with the stated scope.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This markdown file states that rich search results for course lists are available 'auf Englisch verfügbar', which is a natural-language locale restriction. Under the policy, forcing or limiting language/locale without user opt-in or a clearly justified region-specific constraint should be flagged.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

This markdown file contains sample code that sends collected JavaScript error details, including URL, line number, and stack trace, to a remote endpoint via POST. The surrounding text explains the mechanics of remote logging but does not warn readers that such logging can transmit potentially sensitive user or system data and may have privacy implications.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/05.1_webspezifische_leitfaeden_javascript_seo.md (reported line 340)May include surrounding context.

Zu einer URL weiterleiten, unter der vom Server ein 404-Statuscode zurückgegeben wird

text
fetch(`https://api.kitten.club/cats/${id}`)
.then(res => res.json())
.then((cat) => {
if (!cat.exists) {

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/05.1_webspezifische_leitfaeden_javascript_seo.md (reported line 355)May include surrounding context.

Zu einer URL weiterleiten, unter der vom Server ein 404-Statuscode zurückgegeben wird

text
fetch(`https://api.kitten.club/cats/${id}`)
.then(res => res.json())
.then((cat) => {
if (!cat.exists) {

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
80% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/05.1_webspezifische_leitfaeden_javascript_seo.md (reported line 427)May include surrounding context.

md
- Teste deine Seite anschließend noch einmal** mit dem [Test für Rich-Suchergebnisse](https://search.google.com/test/rich-results?hl=de) oder dem [URL-Prüftool](https://search.google.com/search-console?hl=de) in der Search Console.
 Wenn das Problem behoben wurde, siehst du ein grünes Häkchen und es werden keine Fehler angezeigt. Falls weiterhin Fehler angezeigt werden, kannst du einen Beitrag im [Search Central-Hilfeforum](https://support.google.com/webmasters/community?hl=de) posten.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
80% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/05.2_webspezifische_leitfaeden_ecommerce.md (reported line 946)May include surrounding context.

md
- Teste deine Seite anschließend noch einmal** mit dem [Test für Rich-Suchergebnisse](https://search.google.com/test/rich-results?hl=de) oder dem [URL-Prüftool](https://search.google.com/search-console?hl=de) in der Search Console.
 Wenn das Problem behoben wurde, siehst du ein grünes Häkchen und es werden keine Fehler angezeigt. Falls weiterhin Fehler angezeigt werden, kannst du einen Beitrag im [Search Central-Hilfeforum](https://support.google.com/webmasters/community?hl=de) posten.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
80% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/05.1_webspezifische_leitfaeden_javascript_seo.md (reported line 555)May include surrounding context.

md
Solltest du vollkommen unterschiedliche Inhalte für Nutzer und Crawler bereitstellen, kann dies als Cloaking betrachtet werden.
 Das ist beispielsweise der Fall, wenn eine Website für Nutzer eine Seite mit Katzen, für Crawler jedoch eine mit Hunden bereitstellt.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
80% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/05.1_webspezifische_leitfaeden_javascript_seo.md (reported line 687)May include surrounding context.

md
Solltest du vollkommen unterschiedliche Inhalte für Nutzer und Crawler bereitstellen, kann dies als Cloaking betrachtet werden.
 Das ist beispielsweise der Fall, wenn eine Website für Nutzer eine Seite mit Katzen, für Crawler jedoch eine mit Hunden bereitstellt.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
80% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/05.2_webspezifische_leitfaeden_ecommerce.md (reported line 114)May include surrounding context.

md
Hier lernst du gängige UX-Muster für E-Commerce-Websites kennen und erfährst, wie sich diese Muster auf die Fähigkeit von Google auswirken, Inhalte zu crawlen und zu indizieren.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
80% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/05.2_webspezifische_leitfaeden_ecommerce.md (reported line 552)May include surrounding context.

md
Hier lernst du gängige UX-Muster für E-Commerce-Websites kennen und erfährst, wie sich diese Muster auf die Fähigkeit von Google auswirken, Inhalte zu crawlen und zu indizieren.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
80% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/05.2_webspezifische_leitfaeden_ecommerce.md (reported line 1178)May include surrounding context.

md
Wenn deine Website Seiten enthält, auf denen es hauptsächlich um einzelne Videos geht, kann das richtige Markup für vorab aufgenommene Videos (z. B. auf einer Produktseite) oder Livestream-Veranstaltungen Google helfen, die Videos in den Google-Suchergebnissen richtig darzustellen. Weitere Informationen findest du in unserer
 [Dokumentation zum Videoschema-Markup](https://developers.google.com/search/docs/appearance/structured-data/video?hl=de).

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
80% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/05.3_webspezifische_leitfaeden_international.md (reported line 52)May include surrounding context.

md
[Wie Google Seiten crawlt, die an das Gebietsschema angepasst sind](https://developers.google.com/search/docs/specialty/international/locale-adaptive-pages?hl=de)
 Wenn deine Website an das Gebietsschema angepasste Seiten enthält, die basierend auf dem erkannten Land oder der bevorzugten Sprache des Besuchers unterschiedliche Inhalte zurückgeben, crawlt oder indexiert Google möglicherweise nicht alle deine Inhalte für verschiedene Gebiete oder weist nicht allen einen Rang zu.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The document links to Search Central with the explicit hl=de locale and the entire skill content is presented in German, with no indication that users may choose another language or locale. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation unless clearly justified as region-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The changelog repeatedly labels major content additions as '(DE)', indicating a language constraint in the skill materials. Because this natural-language file does not mention user choice, opt-in, or a justified region-specific requirement, it suggests a potential language/locale policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

This markdown file contains user-facing instructional content exclusively in German, and there is no indication that the skill supports other languages or that German is a required locale. Under the policy rule for natural-language violations, forcing a specific language without user opt-in can be a locale-policy issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

Line L0005 states "Sprache: Deutsch," which imposes a specific language in the skill content. Under the policy, language constraints should either be optional for the user or clearly justified as region-specific; this file does not provide such an opt-in or justification.

Content

No source excerpt is available for this finding.