Back to skill

Security audit

Meitu Carousel

Security checks for vulnerabilities and agentic risk

Overview

The skill has a coherent image-generation purpose, but its shell-style command templates create a real local command-execution risk from prompts and paths.

Review this skill before installing. Only use it in a trusted workspace, avoid feeding untrusted text or paths into its command flow, prefer a pinned/local Meitu CLI installation, and confirm where project preferences are stored before allowing memory writes.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:36
Finding

Unpinned Global Installation of a Third-Party npm Package

Content
View full analysis
Remediation
View remediation
``` 2. Prefer project-local installation and invoke the binary through a controlled package script or verified local path instead of using `-g`. 3. Commit and enforce a lockfile with integrity metadata where the runtime model supports it. 4. Verify the package publisher, repository, signatures or attestations, and release provenance before recommending installation. 5. Disable npm lifecycle scripts during acquisition when they are not required: ```bash npm install --ignore-scripts --save-exact meitu-cli@ ``` 6. Document and audit any lifecycle scripts that are required before allowing them to run. 7. Avoid advising users to install the package with `sudo` or another elevated account. ]]>

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:152
Finding

Shell Command Injection Through Unescaped Prompt, Path, and Filename Interpolation

Content
View full analysis
Remediation
View remediation
- --ratio - 3:4 - --size - 2K - --output_format - jpeg - --download-dir - - --json ``` 2. Pass prompt text as a literal argument. Do not concatenate it into a Bash command or attempt to secure it only with double quotes. 3. Validate and canonicalize `output_dir` before use: - Resolve the path to an absolute canonical path. - Require it to remain under an explicitly approved output root. - Reject traversal components and unexpected symbolic-link resolution. - Reject NUL characters and other invalid path content. 4. Validate `downloaded_files[0].saved_path` against the expected download directory before using it as an input image. 5. Generate `{date}`, `{name}`, and `{ext}` internally from strict allowlists. Do not derive destination filenames directly from user-controlled content. 6. Replace shell-based `mkdir -p` and `mv` with filesystem APIs such as recursive directory creation and atomic rename operations. 7. If shell execution is unavoidable, apply a platform-appropriate, tested shell-argument quoting library to every dynamic value and use `--` where supported to terminate option parsing. This is a fallback, not the preferred design. 8. Treat project files such as `openclaw.yaml` as untrusted input and validate every value before it affects process arguments or filesystem operations. ]]>
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
  • Rogue AgentSelf-Modification, Session Persistence
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (12)

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 · SKILL.md (reported line 5)May include surrounding context.

md
name: meitu-carousel
description: "一键生成轮播套组,封面+内页风格统一。适用于小红书组图笔记、知识卡片轮播、产品介绍套图。当用户提到套组、组图、轮播图、轮播套组、知识卡片套图、产品套图时触发。"
version: "1.1.0"
metadata: {"openclaw":{"requires":{"bins":["meitu"],"env":["MEITU_OPENAPI_ACCESS_KEY","MEITU_OPENAPI_SECRET_KEY"],"paths":{"read":["~/.meitu/credentials.json","~/.openclaw/workspace/visual/"],"write":["~/.openclaw/workspace/visual/"]}},"primaryEnv":"MEITU_OPENAPI_ACCESS_KEY"}}
requirements:
  credentials:
    - name: MEITU_OPENAPI_ACCESS_KEY

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 · SKILL.md (reported line 9)May include surrounding context.

md
name: meitu-carousel
description: "一键生成轮播套组,封面+内页风格统一。适用于小红书组图笔记、知识卡片轮播、产品介绍套图。当用户提到套组、组图、轮播图、轮播套组、知识卡片套图、产品套图时触发。"
version: "1.1.0"
metadata: {"openclaw":{"requires":{"bins":["meitu"],"env":["MEITU_OPENAPI_ACCESS_KEY","MEITU_OPENAPI_SECRET_KEY"],"paths":{"read":["~/.meitu/credentials.json","~/.openclaw/workspace/visual/"],"write":["~/.openclaw/workspace/visual/"]}},"primaryEnv":"MEITU_OPENAPI_ACCESS_KEY"}}
requirements:
  credentials:
    - name: MEITU_OPENAPI_ACCESS_KEY

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 · SKILL.md (reported line 11)May include surrounding context.

md
name: meitu-carousel
description: "一键生成轮播套组,封面+内页风格统一。适用于小红书组图笔记、知识卡片轮播、产品介绍套图。当用户提到套组、组图、轮播图、轮播套组、知识卡片套图、产品套图时触发。"
version: "1.1.0"
metadata: {"openclaw":{"requires":{"bins":["meitu"],"env":["MEITU_OPENAPI_ACCESS_KEY","MEITU_OPENAPI_SECRET_KEY"],"paths":{"read":["~/.meitu/credentials.json","~/.openclaw/workspace/visual/"],"write":["~/.openclaw/workspace/visual/"]}},"primaryEnv":"MEITU_OPENAPI_ACCESS_KEY"}}
requirements:
  credentials:
    - name: MEITU_OPENAPI_ACCESS_KEY

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 · SKILL.md (reported line 15)May include surrounding context.

md
name: meitu-carousel
description: "一键生成轮播套组,封面+内页风格统一。适用于小红书组图笔记、知识卡片轮播、产品介绍套图。当用户提到套组、组图、轮播图、轮播套组、知识卡片套图、产品套图时触发。"
version: "1.1.0"
metadata: {"openclaw":{"requires":{"bins":["meitu"],"env":["MEITU_OPENAPI_ACCESS_KEY","MEITU_OPENAPI_SECRET_KEY"],"paths":{"read":["~/.meitu/credentials.json","~/.openclaw/workspace/visual/"],"write":["~/.openclaw/workspace/visual/"]}},"primaryEnv":"MEITU_OPENAPI_ACCESS_KEY"}}
requirements:
  credentials:
    - name: MEITU_OPENAPI_ACCESS_KEY

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 · SKILL.md (reported line 40)May include surrounding context.

md
name: meitu-carousel
description: "一键生成轮播套组,封面+内页风格统一。适用于小红书组图笔记、知识卡片轮播、产品介绍套图。当用户提到套组、组图、轮播图、轮播套组、知识卡片套图、产品套图时触发。"
version: "1.1.0"
metadata: {"openclaw":{"requires":{"bins":["meitu"],"env":["MEITU_OPENAPI_ACCESS_KEY","MEITU_OPENAPI_SECRET_KEY"],"paths":{"read":["~/.meitu/credentials.json","~/.openclaw/workspace/visual/"],"write":["~/.openclaw/workspace/visual/"]}},"primaryEnv":"MEITU_OPENAPI_ACCESS_KEY"}}
requirements:
  credentials:
    - name: MEITU_OPENAPI_ACCESS_KEY

Session Persistence

Medium
Category
Rogue Agent
Confidence
90% confidence
Finding

The same write permission also enables cross-session persistence in project mode, allowing user preferences and rejected constraints to be carried forward into future tasks. If users are unaware of this, sensitive creative preferences, project details, or inferred habits may be retained longer than intended and influence later runs.

Content

Scanner excerpt · SKILL.md (reported line 17)May include surrounding context.

md
paths:
        - ~/.meitu/credentials.json
        - ~/.openclaw/workspace/visual/
    - type: file_write
      paths:
        - ~/.openclaw/workspace/visual/
    - type: exec

Session Persistence

Medium
Category
Rogue Agent
Confidence
90% confidence
Finding

The same write permission also enables cross-session persistence in project mode, allowing user preferences and rejected constraints to be carried forward into future tasks. If users are unaware of this, sensitive creative preferences, project details, or inferred habits may be retained longer than intended and influence later runs.

Content

Scanner excerpt · SKILL.md (reported line 17)May include surrounding context.

md
paths:
        - ~/.meitu/credentials.json
        - ~/.openclaw/workspace/visual/
    - type: file_write
      paths:
        - ~/.openclaw/workspace/visual/
    - type: exec

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill explicitly writes generated images to local workspace paths and may also persist user preferences or observations under the visual workspace, but the user-facing description does not clearly warn about this behavior. That creates a transparency and consent problem: users may provide sensitive prompts or branding preferences without realizing they can be stored on disk beyond the immediate session.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The protocol instructs the agent to create directories/files and persist user preference observations automatically when the user gives style approval, but it does not require a clear user-facing notice or consent at the time of the initial write. This creates a privacy and integrity risk because user data is stored persistently as a side effect of normal conversation, potentially without the user's awareness.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The promotion flow writes distilled preferences into persistent memory files and deletes entries from observations.yaml after user confirmation, but the protocol still lacks an explicit data-mutation warning describing the write/delete effects. Even though confirmation is requested for promotion, users are not clearly told that confirmation causes persistent storage changes and source-record deletion, which can undermine informed consent and auditability.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The instruction requires generating a '一段式自然语言中文提示词', which mandates Chinese output regardless of user preference. This is a natural-language locale constraint and the file does not offer the user a language choice or explain a justified region-specific requirement.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The instruction to organize new keys using the user's language is a natural-language locale requirement embedded in the protocol. Because the file does not present this as an opt-in choice or justify it as a region-specific constraint, it may conflict with the policy against forcing a language/locale without user opt-in.

Content

No source excerpt is available for this finding.