Back to skill

Security audit

Kleinanzeigen.de Helper

Security checks for vulnerabilities and agentic risk

Overview

The skill has a coherent marketplace-listing purpose, but it tells the agent to collect the user's password and SMS login code in chat.

Install only if you are comfortable with the Review risk. Do not send passwords, one-time SMS codes, recovery codes, or session tokens to an agent; a safer version should pause at login and let you type credentials directly into the Kleinanzeigen browser page, then resume only after authentication succeeds.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:50
Finding

Insecure Collection of Account Credentials and SMS Authentication Codes

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 50-52
Vulnerability Type: Plaintext sensitive-data handling
Risk Level: High

Vulnerable Code Snippet:

markdown
2. Frage den Human nach E-Mail + Passwort für den Login
3. Gib die Anmeldedaten ein
4. **SMS-Code**: Der Human gibt den Code, sobald er per SMS ankommt. Warte auf ihn.

English translation:

text
2. Ask the user for their email address and password for login.
3. Enter the login credentials.
4. SMS code: The user provides the code when it arrives by SMS. Wait for it.

Technical Analysis

The skill explicitly instructs the agent to solicit the user's Kleinanzeigen email address, password, and SMS authentication code. Supplying these secrets through the agent exposes them to the conversational context and potentially to model telemetry, logs, session history, memory features, browser automation traces, or other components with access to agent inputs.

This violates secure authentication design and least-exposure principles. Browser automation does not require the agent to know or process the credentials: the workflow can pause while the user enters the password and multifactor authentication code directly into the trusted website interface.

No malicious exfiltration destination is specified, but the insecure handling itself creates an exploitable disclosure channel.

Attack Path

  1. The user invokes the skill to create a Kleinanzeigen listing.
  2. The workflow reaches the login stage.
  3. Following SKILL.md, the agent asks the user to provide an email address and password through the agent conversation.
  4. The agent receives and enters those credentials into the website.
  5. The agent also asks the user to provide the time-sensitive SMS authentication code.
  6. The credentials and MFA code become available in agent-visible context and may be retained in conversation history, telemetry, logs, memory, or tool tr ...[truncated 1035 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove all instructions asking users to send passwords or SMS codes to the agent.
  2. Navigate to the legitimate Kleinanzeigen login page, pause automation, and instruct the user to enter their email address, password, and MFA code directly into the browser.
  3. Resume automation only after the website confirms successful authentication.
  4. Add an explicit warning that passwords, one-time codes, recovery codes, and session tokens must never be entered into chat.
  5. Ensure browser tooling does not capture authentication-field values in screenshots, logs, traces, or action summaries.
  6. Redact sensitive form fields if the browser framework records page state.
  7. Do not store authentication data in skill state, long-term memory, temporary files, or generated listing content.
  8. Prefer an existing authenticated browser session or a platform-supported delegated authorization mechanism where available.
  9. Clear any transient authentication-related context after login and implement retention controls for automation telemetry.
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)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill explicitly instructs the agent to ask for and use the user's email, password, and SMS code, which are highly sensitive authentication factors. This creates a direct credential-harvesting and account-takeover risk, especially because the skill does not require a safer delegated login flow, masking, ephemeral handling, or a warning against storing or reusing the credentials.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The activation condition is overly broad because it triggers whenever the word "kleinanzeigen" appears, even if the user is not asking to create or manage a sales listing. That can cause the agent to invoke browser automation and listing-related flows in the wrong context, increasing the chance of unintended actions or unnecessary collection of sensitive information.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

This markdown file presents all category names in German and does not indicate that the skill is Germany-specific or that users can choose another language. Under the policy rule for language/locale, forcing a specific language without opt-in can be a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

This markdown file contains user-facing instructions and output templates exclusively in German, including the recommended response format. Under the policy for natural-language violations, forcing a specific language without user opt-in can be a locale-policy issue, and no opt-in or justification is provided here.

Content

No source excerpt is available for this finding.