IMA Nano Banana Image Generator

• Outcome: Review
• Latest audit: Sep 11, 2026 · 12:12 PM UTC
• Version: 1.0.9

Overview

The skill is mostly consistent with its image-generation purpose, but failures can persist prompts, image URLs, and local file paths in local logs, so it needs user review before installation.

Install only if you are comfortable sending prompts, image URLs, local image uploads, and your IMA API key to the documented IMA services. Avoid confidential prompts or private image paths unless logs are acceptable on your machine, and periodically delete ~/.openclaw/logs/ima_skills/ and ~/.openclaw/memory/ima_prefs.json if you do not want retained history.

A.I.G

Vulnerability Patterns

• Insecure Skill Coding Practices: Finds exploitable flaws such as hardcoded secrets or command injection
• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally

Findings (1)

T09 · Insecure Skill Coding Practices

Warning

• Location: scripts/ima_image_create.py:747

Finding

Sensitive generation data is persisted in operational logs

Content

View full analysis

Remediation

View remediation

str: parts = urlsplit(value) return urlunsplit((parts.scheme, parts.netloc, parts.path, "", "")) def redact_path(value: str) -> str: return os.path.basename(value) ``` 4. Ensure `extract_error_info()` and retry diagnostics retain structured, sanitized error fields rather than the original exception string. 5. Remove `error_info["message"]` from `attempt_log`, or replace it with a bounded and redacted error category. 6. Make verbose diagnostic logging opt-in rather than enabled by default. Keep production logs at a minimal level. 7. Create the log directory and files with restrictive permissions, such as directory mode `0700` and file mode `0600`, and verify permissions after handler creation. 8. Reduce retention where practical and provide a documented mechanism to disable file logging or immediately delete logs containing user content. 9. Add automated tests asserting that representative secrets, prompts, query parameters, local paths, and image URLs never appear in log output during failed requests and exhausted retries. ]]>

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Taint Tracking: Direct Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain

Findings (17)

Tainted flow: 'task_id' from os.getenv (line 1675, credential/environment) → requests.post (network output)

Critical

• Category: Data Flow
• Confidence: 90% confidence

Finding

Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.

Content

Scanner excerpt · scripts/ima_image_create.py (reported line 803)May include surrounding context.

python

"Check the IMA dashboard for status."
            )

        resp = requests.post(url, json={"task_id": task_id},
                             headers=headers, timeout=30)
        resp.raise_for_status()
        data = resp.json()

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 96% confidence

Finding

The skill description understates important behaviors: sending data to a second domain, persisting user state, and using additional authentication/signing flows. That creates a transparency and consent problem, and could cause users or reviewers to approve a skill without understanding where credentials and local images are sent or what state is retained.

Content

No source excerpt is available for this finding.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 99% confidence

Finding

The skill description understates important behaviors: sending data to a second domain, persisting user state, and using additional authentication/signing flows. That creates a transparency and consent problem, and could cause users or reviewers to approve a skill without understanding where credentials and local images are sent or what state is retained.

Content

No source excerpt is available for this finding.

Credential Access

High

• Category: Privilege Escalation
• Confidence: 70% confidence

Finding

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

Content

Scanner excerpt · scripts/ima_image_create.py (reported line 1557)May include surrounding context.

python

args   = build_parser().parse_args()
    base   = DEFAULT_BASE_URL
    
    # Get API key from args or environment variable
    apikey = args.api_key or os.getenv("IMA_API_KEY")
    if not apikey:
        logger.error("API key is required. Use --api-key or set IMA_API_KEY environment variable")

Undeclared Tool Scope

Medium

• Category: MCP Least Privilege
• Confidence: 89% confidence

Finding

The skill declares sensitive capabilities via metadata and behavior (environment credential use, local file writes, and outbound network access) but does not define an explicit tool/permission scope. That weakens sandboxing and reviewability, making it easier for an agent runtime to grant broader access than users expect.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

• Category: Not specified by scanner
• Confidence: 95% confidence

Finding

The mandatory Step 1-4 notification templates are written only in Chinese, while the document instructs the agent to use this messaging flow authoritatively. Although Step 0 mentions replying in the user's language, the required later steps do not provide an opt-in or explicit alternative, which can violate language/locale policy by forcing a specific language.

Content

No source excerpt is available for this finding.

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 · scripts/ima_image_create.py (reported line 741)May include surrounding context.

python

f"credit={credit}, attribute_id={attribute_id}")

    try:
        resp = requests.post(url, json=payload, headers=headers, timeout=30)
        resp.raise_for_status()
        data = resp.json()

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 · scripts/ima_image_create.py (reported line 803)May include surrounding context.

python

f"credit={credit}, attribute_id={attribute_id}")

    try:
        resp = requests.post(url, json=payload, headers=headers, timeout=30)
        resp.raise_for_status()
        data = resp.json()

Missing User Warnings

Medium

• Category: Not specified by scanner
• Confidence: 87% confidence

Finding

The script writes per-user preference data to local disk without any user-facing disclosure or consent in this file. While the data is not highly sensitive, undisclosed persistence can create privacy risk, especially on shared systems or multi-user agent environments where local state may be accessible to others.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium

• Category: Not specified by scanner
• Confidence: 84% confidence

Finding

The manifest describes a narrowly scoped image-generation skill, but this module creates a dedicated directory in ~/.openclaw and persists logs there. Local file creation and retention are not mentioned in the manifest description, so the implemented behavior exceeds the stated scope at the description level.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium

• Category: Not specified by scanner
• Confidence: 86% confidence

Finding

The manifest presents the skill as a Nano Banana image generator, but this code also removes files from the user's home-directory log store based on age. File deletion is a side effect beyond straightforward image generation and is not disclosed by the manifest description.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low

• Category: Not specified by scanner
• Confidence: 92% confidence

Finding

The manifest description contains both Chinese and English text, which implies a language presentation choice but does not indicate whether users can select their preferred language. Under the policy criteria, forcing or assuming a language/locale without opt-in can be a natural-language policy issue.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low

• Category: Supply Chain
• Confidence: 96% confidence

Finding

The dependency is specified as requests>=2.25.0, which allows any future version and does not guarantee a reproducible or reviewed install. This creates supply-chain risk and makes it possible to unintentionally install a vulnerable or incompatible release, especially in automation or fresh environments.

Content

Scanner excerpt · requirements.txt (reported line 4)May include surrounding context.

text

# Python dependencies for ai-nano-banana-ima (Nano Banana image generation)
# Install with: pip install -r requirements.txt

requests>=2.25.0

Unverifiable Dependency: requests has 16 known advisory(ies) (CVE-2014-1830 (Exposure of Sensitive Information to an Unauthorized Actor in Requests); CVE-2024-47081 (Requests vulnerable to .netrc credentials leak via malicious URLs); CVE-2024-35195 (Requests `Session` object does not verify requests after making first request wi) +13 more), but the manifest does not pin a version, so it is unknown whether the installed release is affected

Low

• Category: Supply Chain
• Confidence: 91% confidence

Finding

requests has multiple known advisories, and because the manifest does not pin a specific version, there is no way to verify whether installed environments will receive a fixed or vulnerable release. In a skill that interacts with external APIs using an API key, uncertainty around HTTP client security increases the risk of credential leakage or unsafe request handling if an affected version is resolved.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low

• Category: Not specified by scanner
• Confidence: 95% confidence

Finding

The header builder defaults language to "en" and always sends x_app_language with that value. This imposes a specific language/locale unless the caller explicitly overrides it, which is a natural-language policy concern because there is no opt-in before choosing English as the default.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low

• Category: Not specified by scanner
• Confidence: 90% confidence

Finding

The manifest describes an image generation skill focused on IMA Open API operations, but this file also implements a user preference memory subsystem that reads and writes ~/.openclaw/memory/ima_prefs.json. Persisting user-specific history is not disclosed in the manifest description and goes beyond the core generation/polling workflow.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low

• Category: Not specified by scanner
• Confidence: 82% confidence

Finding

Saving model usage by user ID to a local memory file introduces stateful tracking behavior unrelated to the manifest's stated role of generating images via the IMA API. This is not an obvious requirement for text_to_image or image_to_image execution itself.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.