Back to skill

Security audit

Image Paginator(律师证据长截图转PDF)

Security checks for vulnerabilities and agentic risk

Overview

This skill appears to do local image-to-PDF conversion, but its installer fetches mutable remote code and unpinned Python packages, so users should review it before installing.

Review before installing. Prefer installing from a pinned release or reviewed local copy, avoid running install.sh as an administrator, use an isolated virtual environment with pinned dependencies, and ask the agent not to automatically open generated PDFs unless you explicitly want that.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
Findings (4)

T03 · Remote Payload Retrieval and Execution

Error
Location
install.sh:9
Finding

Mutable Remote Payload Is Downloaded and Installed Without Integrity Verification

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
install.sh:31
Finding

Unpinned Python Packages Are Installed Into the Invoking Environment

Content
View full analysis
/dev/null; then pip3 install --quiet fpdf2 Pillow elif command -v pip &>/dev/null; then pip install --quiet fpdf2 Pillow else echo " ⚠️ 未找到 pip3,请手动运行: pip3 install fpdf2 Pillow" fi ``` ### Technical Analysis The installer resolves and installs the latest available versions of `fpdf2` and `Pillow` at installation time. No version constraints, hashes, lock file, trusted package index policy, or signature verification are used. The command also invokes whichever `pip3` or `pip` executable appears first in `PATH`. It does not create a dedicated virtual environment, so packages may be installed into or modify the user's existing Python environment. The effective dependency graph can consequently change after the Skill has been reviewed. This does not establish that either named package is malicious. The risk comes from mutable dependency resolution and the absence of integrity and environment-isolation controls. ### Attack Path 1. A dependency account, release pipeline, package repository, or transitive dependency is compromised. 2. A malicious or otherwise unsafe package version becomes the version selected by `pip`. 3. A user runs `install.sh`. 4. `pip` downloads the newly resolved package and runs any package installation or build behavior. 5. The package executes with the privileges of the user running the installer and remains available in the affected Python environment. An additional local attack path exists if an attacker can influence the installer's `PATH`: a spoofed `pip` or `pip3` executable could be selected and run. ### Impact Assessment A compromised dependency or spoofed package-management executable can execute code with the invok ...[truncated 407 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/slice_n_pdf.py:176
Finding

Predictable Shared Temporary Files Expose Sensitive Image Data to Local Races

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/slice_n_pdf.py:256
Finding

Cleanup Recursively Deletes a Pre-Existing Destination Directory

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
Findings (12)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The declared purpose is local image-to-PDF conversion, but the skill also includes downloading code from GitHub, unpacking into ~/.qclaw/skills/, and installing packages with pip3. That mismatch is dangerous because agents or users may trust the skill as a simple local formatter while it performs supply-chain and system-modifying operations outside the stated scope.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill explicitly instructs use of shell commands (python, open, bash, curl, unzip, pip3) but does not declare any tool scope or allowed-tools metadata. This creates an authorization and transparency gap: an agent may execute filesystem and process-launching actions that users or the platform did not clearly consent to.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The guardrail tells the agent to always run open <path> after generation, which launches a local application and causes side effects beyond the core task of producing a PDF. Automatic app launching can expose sensitive documents on screen, trigger risky file handlers, or violate user expectations in unattended environments.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill directs the agent to open the generated PDF without warning or confirmation. Forced execution of a viewer command is a security-relevant side effect because it bypasses user consent and may reveal privileged legal evidence or trigger downstream application behavior.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
98% confidence
Finding

The installation instructions download an archive from GitHub via curl, unzip it into a live skills directory, and install Python packages with pip3. This is a classic supply-chain risk: remote content may change, be tampered with, or introduce malicious code, and the writes/installations affect the local environment beyond the user's immediate conversion task.

Content

Scanner excerpt · SKILL.md (reported line 171)May include surrounding context.

bash install.sh

方法二:直接下载

curl -fsSL https://github.com/wux818738-alt/openclaw-skill-image-paginator/archive/refs/heads/main.zip -o /tmp/skill.zip unzip /tmp/skill.zip -d ~/.qclaw/skills/ pip3 install fpdf2 Pillow

text

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The installer fetches code from a remote GitHub ZIP at install time and also installs Python packages from the network without pinning versions, integrity verification, or any trust prompt. This creates a supply-chain risk: if the repository, transport path, or dependency resolution is compromised, arbitrary code could be installed under the user's account.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The script unconditionally removes the existing skill directory with rm -rf and does not ask for confirmation or create a backup. While this is not code execution, it can cause unintended data loss or destroy local modifications if the path already contains user data.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
88% confidence
Finding

This curl command pulls executable installer content from an external source, which is risky because the downloaded archive is later unpacked and installed without authenticity or integrity checks. In the context of a simple image-to-PDF skill, network retrieval is operationally understandable for installation, but it still expands the attack surface and enables supply-chain compromise.

Content

Scanner excerpt · install.sh (reported line 23)May include surrounding context.

sh
# 下载 zip
echo "   ⬇️  下载: $ZIP_URL"
curl -fsSL "$ZIP_URL" -o "$ZIP_FILE"

# 解压(自动解压到 image-paginator-main/)
UNZIP_DIR=$(mktemp -d)

Dynamic import via __import__()

Medium
Category
Dangerous Code Execution
Confidence
75% confidence
Finding

Dynamic import() can load arbitrary modules at runtime, bypassing static analysis and potentially importing malicious code.

Content

Scanner excerpt · scripts/slice_n_pdf.py (reported line 206)May include surrounding context.

python
# ── CLI ───────────────────────────────────────────────────────
def main():
    ap = (
        __import__("argparse")
        .ArgumentParser(
            description="Tile long image(s) into a numbered grid PDF.",
        )

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

Natural-language policy violations include forcing a specific language or locale without user opt-in. The description and major workflow sections present the skill in Chinese for a specific audience, but the file does not explicitly offer the user a language choice or document that the locale constraint is required.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

All user-facing messages, comments describing purpose, and usage examples are presented only in Chinese. This imposes a single language on users without an opt-in or alternative, which can violate language/locale policy when the skill is not clearly documented as region-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The inline comment 参数校验 is written only in Chinese, introducing a language-specific instruction/context in a file that otherwise uses English and does not state any locale choice or justification. This can violate a language/locale policy that requires neutral language or user opt-in for a specific locale.

Content

No source excerpt is available for this finding.