Back to skill

Security audit

realsense-d435i

Security checks for vulnerabilities and agentic risk

Overview

This appears to be a real RealSense camera capture tool, but it needs review because it records sensitive sensor data locally and has unsafe FFmpeg path handling.

Install only if you are comfortable with a script that can access a RealSense camera and save RGB, depth, point-cloud, video, and IMU data to disk. Use a virtual environment, avoid administrative privileges, use a simple trusted output path without quotes or newlines, and review or patch the FFmpeg manifest handling before using RGBD recording in shared or untrusted workflows.

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

Note
Location
SKILL.md:17
Finding

Unpinned Third-Party Python Dependencies

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/realsense_capture.py:174
Finding

FFmpeg Concat Manifest Injection Through User-Controlled Output Path

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (9)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding

The skill documents shell commands and file-writing behavior but does not declare any tool scope such as allowed-tools or permissions. That creates an authorization/expectation gap: an agent or user may invoke shell and write-to-disk behavior without explicit consent boundaries, increasing the chance of unintended command execution or data persistence.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/realsense_capture.py (reported line 83)May include surrounding context.

python
print("[PointCloud] Trying rs-pointcloud utility...")
    ply_path = out_path("pointcloud", "ply", "pointcloud")
    try:
        r = subprocess.run(["rs-pointcloud", "--output", ply_path,
                            "--format", "ply"],
                           capture_output=True, text=True, timeout=5)
        if r.returncode == 0:

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The manifest describes a RealSense capture skill for depth photos, point clouds, IMU data, and RGBD recording, which justifies camera access and local file output. However, the implementation also depends on spawning external executables via subprocess, introducing broader host-execution capability not explicitly justified by that stated purpose.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/realsense_capture.py (reported line 205)May include surrounding context.

python
with open(rgb_list, "w") as f:
        for fp in rgb_frames: f.write(f"file '{fp}'\n")
    subprocess.run(["ffmpeg", "-y", "-f", "concat", "-safe", "0", "-i", rgb_list,
                   "-c:v", "libx264", "-preset", "ultrafast", "-crf", "23", rgb_out],
                  capture_output=True)

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Recording RGBD video is within the manifest's stated purpose, but the manifest does not indicate that the skill will execute system-level encoding tools. This expands the capability surface from device capture and file storage to launching external programs on the host.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/realsense_capture.py (reported line 211)May include surrounding context.

python
with open(depth_list, "w") as f:
        for fp in depth_frames: f.write(f"file '{fp}'\n")
    subprocess.run(["ffmpeg", "-y", "-f", "concat", "-safe", "0", "-i", depth_list,
                   "-c:v", "libx264", "-preset", "ultrafast", "-crf", "23", depth_out],
                  capture_output=True)

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The manifest description and main body are written in Chinese, with no indication that users may choose another language or that the skill is intentionally restricted to a Chinese-speaking audience. Under the stated policy, forcing a specific language without opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill captures and stores RGB images, depth data, point clouds, video, and IMU telemetry to disk, but the description does not present a prominent user warning about recording and persistence. This can lead to inadvertent collection of sensitive environmental imagery or motion data, especially in private or regulated spaces.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file includes user-facing instructional text in Chinese for the default output behavior, while the rest of the script is primarily in English. This imposes a specific language on users without opt-in or a documented locale-specific reason, which matches the language/locale policy concern.

Content

No source excerpt is available for this finding.