Back to skill

Security audit

Roblox Clip Transformer

Security checks for vulnerabilities and agentic risk

Overview

This Roblox video helper is mostly purpose-aligned, but it needs Review because its FFmpeg wrappers have unsafe text/path handling, silent overwrite behavior, and documentation that overstates implemented editing features.

Install only if you are comfortable running local FFmpeg/Python media scripts from this package. Use trusted input values for titles, CTAs, subtitle paths, music paths, and output paths; avoid letting untrusted text or filenames flow into the rendering commands. Expect existing output files to be replaced, and pin/review Python dependencies before use in a production or team environment.

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 (3)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/platform-render.py:216
Finding

Unescaped Title and CTA Values Permit FFmpeg Filter-Graph Injection

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/caption-sync.py:107
Finding

Incomplete Subtitle-Path Escaping Permits FFmpeg Filter Injection

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Note
Location
SKILL.md:124
Finding

Unpinned Third-Party Dependencies Create a Mutable Supply-Chain Boundary

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (19)

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The declared description promises an end-to-end promotional video creation/editing tool, but the supplied code only performs footage analysis. It reads a local video file, invokes ffprobe/ffmpeg to detect scene changes, silence, and audio levels, then writes analysis results and recommendations to JSON. While this analysis could support smart trimming, it does not implement the core advertised capabilities of producing edited TikTok/Shorts/Reels content, adding captions, syncing music, converting aspect ratios, or applying overlays. Therefore the actual behavior is materially narrower and different from the declared primary purpose.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The declared description suggests a broad end-to-end promotional video editing pipeline for Roblox gameplay footage. However, the supplied code chunk implements only one narrow feature: generating synchronized captions from audio and optionally embedding them into the video. There is no evidence of clip selection, trimming, reframing for TikTok/Shorts/Reels, music synchronization, branding overlays, or general promotional content assembly. While captions are one part of the declared feature set, the code’s primary purpose is materially narrower than the declared purpose, so this is a clear description-behavior mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The description claims a broad promotional-video creation tool for Roblox recordings, including multiple concrete editing features. The code chunk only performs music analysis on an input audio file using librosa and outputs beat/peak metadata plus suggested cut points. Although music synchronization is mentioned in the description, even the video sync function here is a stub and does not actually edit video. This is therefore a material mismatch in primary purpose and implemented capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The code generally aligns with the stated purpose of rendering promotional social video variants from gameplay footage, but there are meaningful description/behavior gaps. Aspect ratio conversion, optional music, and brand/title/CTA overlays are implemented. Analysis-based segment selection exists, but the selected segments are never actually used to cut and concatenate clips in the FFmpeg command, so 'smart trimming' is not truly carried out as described. The description also explicitly includes captions, but the code has no caption-generation or subtitle-overlay implementation; it only supports simple drawtext title/CTA overlays. Finally, the declaration emphasizes TikTok, YouTube Shorts, and Reels, while the code supports a generic 'youtube' 16:9 output rather than a dedicated Shorts mode. These are substantive mismatches in represented capability, though the overall domain and intent are close.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill advertises and demonstrates shell execution and output file creation, but it does not declare any tool scope such as allowed-tools or permissions. In an agent environment, this can lead to overbroad execution authority, making it easier for the skill to invoke shell commands or write files without explicit review or least-privilege constraints.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 74)May include surrounding context.

md
### Smart Trimming

Automatically removes:
- Menu screens
- Loading screens
- AFK/idle moments

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/analyze-footage.py (reported line 29)May include surrounding context.

python
"-show_format", "-show_streams",
        video_path
    ]
    result = subprocess.run(cmd, capture_output=True, text=True)
    if result.returncode != 0:
        raise RuntimeError(f"ffprobe failed: {result.stderr}")
    return json.loads(result.stdout)

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/analyze-footage.py (reported line 42)May include surrounding context.

python
"-show_format", "-show_streams",
        video_path
    ]
    result = subprocess.run(cmd, capture_output=True, text=True)
    if result.returncode != 0:
        raise RuntimeError(f"ffprobe failed: {result.stderr}")
    return json.loads(result.stdout)

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/analyze-footage.py (reported line 63)May include surrounding context.

python
"-show_format", "-show_streams",
        video_path
    ]
    result = subprocess.run(cmd, capture_output=True, text=True)
    if result.returncode != 0:
        raise RuntimeError(f"ffprobe failed: {result.stderr}")
    return json.loads(result.stdout)

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/analyze-footage.py (reported line 84)May include surrounding context.

python
"-show_format", "-show_streams",
        video_path
    ]
    result = subprocess.run(cmd, capture_output=True, text=True)
    if result.returncode != 0:
        raise RuntimeError(f"ffprobe failed: {result.stderr}")
    return json.loads(result.stdout)

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/caption-sync.py (reported line 126)May include surrounding context.

python
"-show_format", "-show_streams",
        video_path
    ]
    result = subprocess.run(cmd, capture_output=True, text=True)
    if result.returncode != 0:
        raise RuntimeError(f"ffprobe failed: {result.stderr}")
    return json.loads(result.stdout)

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/platform-render.py (reported line 109)May include surrounding context.

python
"-show_format", "-show_streams",
        video_path
    ]
    result = subprocess.run(cmd, capture_output=True, text=True)
    if result.returncode != 0:
        raise RuntimeError(f"ffprobe failed: {result.stderr}")
    return json.loads(result.stdout)

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The module docstring says the script uses beat detection 'to synchronize video transitions with music rhythm', implying automated transition syncing. In practice, the code only analyzes audio and emits beat/peak metadata; the only video-sync function is unimplemented and does not perform transitions.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The docstring at L140 says this function will 'Create video synced to beats using FFmpeg', but all transition branches are placeholders with pass statements and no FFmpeg invocation occurs. The function instead prints that beat sync data was saved for manual use and returns True, which actively contradicts the documented behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest advertises 'captions' as part of the automatic editing pipeline, but the code only supports title and CTA text overlays. The build_filter_chain signature includes an add_captions parameter that is never used, and no subtitle/caption input, generation, or FFmpeg subtitle filter is present.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The FFmpeg command is built with the '-y' flag, which forces overwriting of the output file if it already exists. Although the script's purpose is rendering video, there is no confirmation prompt or explicit user-facing warning that an existing output file will be replaced automatically.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
88% confidence
Finding

Although ffmpeg is also invoked via an argument list, this code builds the -vf filter string from untrusted title/cta values without escaping FFmpeg filter syntax. An attacker controlling these fields can break filter parsing or potentially trigger unintended FFmpeg behavior, creating a command-to-interpreter injection issue within FFmpeg's own mini-language even without shell=True.

Content

Scanner excerpt · scripts/platform-render.py (reported line 238)May include surrounding context.

python
if verbose:
        print(f"Running: {' '.join(cmd)}")
    
    result = subprocess.run(cmd, capture_output=True, text=True)
    
    if result.returncode != 0:
        print(f"FFmpeg error: {result.stderr}")

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The script writes analysis results to the user-supplied or auto-derived output path using open(..., 'w'), which can overwrite an existing JSON file. Although writing output is part of the tool's purpose, there is no explicit warning or confirmation when an existing file may be replaced.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The comments explicitly describe a crop-based transformation, yet the FFmpeg filters create a split stream, blur a background layer, scale the foreground with aspect preservation, and overlay it centered. This is an active contradiction between implementation comments and actual behavior, not just missing detail.

Content

No source excerpt is available for this finding.