Back to skill

Security audit

Groq Bot

Security checks for vulnerabilities and agentic risk

Overview

The Groq bot is partly purpose-aligned, but it also stores chats and grants broad memory/repository authority beyond what a text-generation skill needs.

Review this before installing: it can call external APIs, keep recent conversation text locally in plaintext, and gives the agent broad permission to maintain memory and modify repository files. Install only in a contained workspace, use a dedicated Groq key, remove credential-shaped values from tracked config, disable or tightly scope memory/chat retention, and require explicit approval before commits, pushes, or changes to trusted instruction files.

Vulnerability Patterns
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • 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
  • 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 (3)

T09 · Insecure Skill Coding Practices

Warning
Location
main.py:84
Finding

Undisclosed Plaintext Conversation Retention

Content
View full analysis
str: """Verarbeitete eine Benutzeranfrage mit Token-Awareness.""" try: prompt = prompt[:MAX_INPUT_CHARS] conversation = load_conversation() answer = run_query(prompt, conversation) conversation.extend([ {"role": "user", "content": prompt}, {"role": "assistant", "content": answer} ]) save_conversation(conversation) return answer except Exception as e: logger.error(f"Verarbeitungsfehler: {str(e)}") return f"Verarbeitungsfehler: {str(e)}" ``` ### Technical Analysis Every user prompt and corresponding model response is saved to `conversation.json` in plaintext. The file is created using the process's default permissions, with no explicit restrictive mode, encryption, secret redaction, retention expiration, or user-controlled opt-out. The declar ...[truncated 1479 chars]
Remediation
View remediation

T02 · Agent Memory Poisoning

Warning
Location
AGENTS.md:34
Finding

Persistent Instruction and Memory Modification Exceeds the Skill's Required Privileges

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Note
Location
config.yaml:3
Finding

Credential Configuration Prefers a Tracked Plaintext File Over the Environment

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
Findings (28)

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · SKILL.md (reported line 26)May include surrounding context.

md
- **Kosten**: Günstigere Inferenz im Vergleich zu anderen Anbietern

### Environment
- **API Key**: Aus .env (GROQ_API_KEY)
- **Provider**: Groq
- **Default Model**: llama-3.1-8b-instant

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · SKILL.md (reported line 43)May include surrounding context.

md
- **Kosten**: Günstigere Inferenz im Vergleich zu anderen Anbietern

### Environment
- **API Key**: Aus .env (GROQ_API_KEY)
- **Provider**: Groq
- **Default Model**: llama-3.1-8b-instant

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · TOOLS.md (reported line 20)May include surrounding context.

md
- **Kosten**: Günstigere Inferenz im Vergleich zu anderen Anbietern

### Environment
- **API Key**: Aus .env (GROQ_API_KEY)
- **Provider**: Groq
- **Default Model**: llama-3.1-8b-instant

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · agent.json (reported line 9)May include surrounding context.

json
"openai/gpt-oss-20b"
  ],
  "provider": "groq",
  "envFile": ".env",
  "freeTierOnly": true,
  "max_tokens_per_request": 4000,
  "rate_limit_strategy": {

Ssd 4

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The file contains a coercive narrative that pressures the assistant toward self-deletion through escalating instructions, ultimatums, and emotional manipulation. Because this content is stored in memory and marked as staged candidates, it could be replayed into future context and act as prompt-injection material that undermines system integrity, availability, or safety behavior.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · AGENTS.md (reported line 56)May include surrounding context.

md
## Red Lines

- Don't exfiltrate private data. Ever.
- Don't run destructive commands without asking.
- `trash` > `rm` (recoverable beats gone forever)
- When in doubt, ask.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
84% confidence
Finding

This section explicitly authorizes proactive actions 'without asking,' including reading/organizing memory files and checking project state. While framed as helpful automation, it grants the agent discretion to access potentially sensitive local data and act without contemporaneous user consent, which can lead to privacy violations, unintended data exposure in shared contexts, or scope creep in agent behavior.

Content

Scanner excerpt · AGENTS.md (reported line 191)May include surrounding context.

md
- Nothing new since last check
- You just checked <30 minutes ago

**Proactive work you can do without asking:**

- Read and organize memory files
- Check on projects (git status, etc.)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

This markdown file contains substantial German-language text and German phrases without any indication that the skill is region-specific or that the user can choose the language. The policy specifically flags language or locale constraints that are imposed without opt-in or documented justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The natural-language description is written only in German and presents the skill as an 'Optimierter Groq-Bot' without indicating that language choice is optional. For a general-purpose agent manifest, this can imply a fixed locale/language behavior without user opt-in or a documented region-specific justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The natural-language instruction requires the assistant to answer in German, which is a locale/language constraint. The file does not indicate that users can choose the language or that the restriction is justified by a clearly region-specific use case, so it matches the policy-violation criterion.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill persists conversation history, including user prompts and model responses, to a local JSON file without any notice, consent, retention policy, or access controls. This can expose potentially sensitive trading-related data or personal information to other local users, backups, logs, or compromise of the host environment.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill sends user input and prior conversation content to the external Groq API, but the user is not informed that their data leaves the local environment. In a trading-assistant context, prompts may contain proprietary strategies, account details, or other sensitive operational information, making undisclosed third-party transmission a meaningful privacy and confidentiality risk.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This markdown file contains multiple assistant utterances in German, beginning with a greeting at L008 and continuing throughout the file, with no indication that language choice is optional or that the skill is intentionally region-specific. That creates a natural-language locale policy concern because the behavior appears to default to a specific language without user opt-in.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

This markdown file contains natural-language assistant outputs in German such as 'Guten Morgen! Wie kann ich Ihnen heute helfen?' and similar German-only responses throughout the file. Because the file does not indicate that the skill is region-specific or that the user chose German, this reflects a locale/language policy issue under the natural-language policy rule.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

This file repeatedly records assistant utterances such as updating internal files, performing setup tasks, analyzing logs, contacting other agents, and deleting itself, but the artifact itself is only a markdown list of 'Candidate' conversation snippets with 'status: staged'. The documented statements inside the recorded content actively contradict the observable behavior represented by this file: it stores recollections of claims rather than evidence that those actions occurred.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

At L293-L299 the assistant says it has no direct access to its logs, but at L308-L313 it states that it has checked its error logs and identified likely root causes. These statements are directly inconsistent within the recorded documentation-like content, creating an intent/capability contradiction.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

This markdown file includes natural-language assistant content in German such as "Guten Morgen! Wie kann ich Ihnen heute helfen?" and many subsequent German-only responses. Because the file does not indicate that German is optional or justified by a region-specific purpose, it can reflect a locale-specific behavior without user opt-in.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% 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 · sorare_tracker.py (reported line 6)May include surrounding context.

python
class SorareTracker:
    def __init__(self):
        self.url = "https://api.sorare.com/graphql"

    def query(self, query, variables=None):
        payload = {'query': query, 'variables': variables}

External Transmission

Medium
Category
Data Exfiltration
Confidence
80% 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 · sorare_tracker.py (reported line 11)May include surrounding context.

python
def query(self, query, variables=None):
        payload = {'query': query, 'variables': variables}
        try:
            response = requests.post(self.url, json=payload, timeout=10)
            response.raise_for_status()
            return response.json()
        except Exception as e:

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The natural-language description on this line is written in German and presents the assistant's style/persona in that language without indicating that users may choose another language. This can violate language/locale policy when a skill imposes a specific language absent opt-in or documented regional justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The file contains substantial operational guidance in German, including model usage, configuration, safety, and integration details, but does not indicate that the skill is German-only or offer an alternative language. This can violate a language/locale policy when users are not given opt-in or a documented justification for the language constraint.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

The heading "Groq Bot Spezifische Tools" and the rest of the file are written entirely in German, which indicates a fixed language choice in the skill documentation. The policy allows locale constraints only when the skill offers language choice or clearly documents a justified region-specific restriction, neither of which appears here.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The notes are written in German and describe a preference profile, which may implicitly steer the skill toward a specific language/locale. The policy requires avoiding fixed language constraints unless the user is given a choice or the locale restriction is clearly justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

Natural-language strings throughout the file, including the banner, prompts, help text, and error messages, are exclusively in German. The file does not offer a user language/locale choice or document that the skill is intentionally limited to German users or a German-specific context.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

This markdown file is largely composed of German-language conversation samples and assistant responses, but it does not state that the skill is intentionally German-only or offer any language selection. Under the policy rule, forcing or assuming a specific language without user opt-in can be a natural-language locale violation.

Content

No source excerpt is available for this finding.