Back to skill

Security audit

Agnes Image 2.0 Flash

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent image-generation skill, but it sends user prompts and image URLs to a third-party API and then downloads an unvalidated API-supplied URL to a local file.

Install only if you are comfortable sending prompts and public image URLs to Agnes AI. Avoid confidential prompts or private/proprietary images, and prefer using it in a constrained environment because the script saves output from an API-supplied URL without validating the download target or file type.

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

Warning
Location
scripts/agnes-image.sh:110
Finding

Unrestricted Download from a Server-Controlled URL

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • 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
Findings (9)

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · scripts/agnes-image.sh (reported line 84)May include surrounding context.

sh
fi

# Send request
RESPONSE=$(curl -sf --max-time 360 \
  -H "Authorization: Bearer ${API_KEY}" \
  -H "Content-Type: application/json" \
  -d "$BODY" \

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The trigger keywords include broad, everyday phrases such as '合并' and '创建图像' that can match routine user requests outside the intended image-generation context. This increases the chance of accidental skill activation, causing unintended external API use, unnecessary processing, or disclosure of prompts and image inputs to the remote service.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill documents sending prompts and public image URLs to an external image-generation API but does not clearly warn users that their content will be transmitted to a third-party service. In this context, accidental triggering is more dangerous because users may unknowingly expose sensitive prompts, proprietary images, or personal data during generation or editing workflows.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

The curl request transmits constructed request data, including prompt content and image references, to an external endpoint using an API key. While this is core functionality, it is still a true external data exfiltration path from the local environment to a remote service, which is security-relevant if users may pass confidential data.

Content

Scanner excerpt · scripts/agnes-image.sh (reported line 84)May include surrounding context.

sh
fi

# Send request
RESPONSE=$(curl -sf --max-time 360 \
  -H "Authorization: Bearer ${API_KEY}" \
  -H "Content-Type: application/json" \
  -d "$BODY" \

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The script sends the user-supplied prompt and any provided image URLs to a third-party API service without any explicit notice, consent check, or data-classification guardrail. In an agent-skill context, prompts and URLs may contain sensitive internal data, making unintended external disclosure a real privacy and compliance risk.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
70% 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 · scripts/test-connection.sh (reported line 24)May include surrounding context.

sh
# Test with a simple request
echo "📡 Sending test request..."
RESPONSE=$(curl -sf --max-time 30 \
  -H "Authorization: Bearer ${API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

This markdown file presents headings, instructions, and best practices in Chinese throughout, but does not indicate that the language is optional or limited to a specific regional audience. Under the policy rule for natural-language violations, forcing a specific language without user opt-in is reportable.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
79% confidence
Finding

This markdown file is primarily presented for Chinese-speaking users, but nearly all example prompts are fixed in English and the document does not indicate that users may choose their preferred language. Because the policy scope includes language or locale constraints in natural-language content, this can be treated as a mild language-policy issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The entire guide is written in Chinese and does not indicate that other languages are supported or that the user may choose a preferred locale. Under the policy for natural-language violations, forcing a specific language without opt-in can be considered a locale-policy issue.

Content

No source excerpt is available for this finding.