Back to skill

Security audit

CC Video Creation

Security checks for vulnerabilities and agentic risk

Overview

This video-production skill mostly does what it says, but it under-discloses external voice-generation data sharing and has local file-handling risks users should review first.

Install only if you are comfortable with narration text being sent to Edge TTS for voice generation, and avoid putting secrets or private drafts in scripts. Use an authorized Kling AI account rather than relying on the published email, and prefer fixing the shared /tmp frame directory before running this on a multi-user system.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T09 · Insecure Skill Coding Practices

Warning
Location
cc_animator.py:21
Finding

Predictable Shared Temporary Directory Enables File Overwrite and Deletion

Content
View full analysis
Remediation
View remediation

other

Note
Location
scripts/produce_animated.py:298
Finding

Undisclosed External Transmission of Complete Narration Text Through Edge TTS

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (36)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill claims 'no API' and broad support for Manim/Kling/upload, yet the apparent implementation uses different tooling and external service dependencies such as Edge TTS. Misrepresenting external dependencies and capabilities is security-relevant because it can conceal network use, third-party data exposure, or additional trust boundaries from reviewers.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill claims 'no API' and broad support for Manim/Kling/upload, yet the apparent implementation uses different tooling and external service dependencies such as Edge TTS. Misrepresenting external dependencies and capabilities is security-relevant because it can conceal network use, third-party data exposure, or additional trust boundaries from reviewers.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

The skill claims 'no API' and broad support for Manim/Kling/upload, yet the apparent implementation uses different tooling and external service dependencies such as Edge TTS. Misrepresenting external dependencies and capabilities is security-relevant because it can conceal network use, third-party data exposure, or additional trust boundaries from reviewers.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

The skill claims 'no API' and broad support for Manim/Kling/upload, yet the apparent implementation uses different tooling and external service dependencies such as Edge TTS. Misrepresenting external dependencies and capabilities is security-relevant because it can conceal network use, third-party data exposure, or additional trust boundaries from reviewers.

Content

No source excerpt is available for this finding.

Unvalidated Output Injection

High
Category
Output Handling
Confidence
95% confidence
Finding

Model output is used without validation or sanitization. Unvalidated output injected into downstream contexts (SQL, shell, HTML) enables injection attacks and arbitrary code execution.

Content

Scanner excerpt · cc_illustrator.py (reported line 611)May include surrounding context.

python
def render_video(name="cc-illustrated-09", fps=30):
    print("\n🎞 Renderer video...")
    output = OUTPUT / f"{name}.mp4"
    result = subprocess.run([
        "ffmpeg", "-y", "-framerate", str(fps),
        "-i", str(FRAMES / "frame_%05d.png"),
        "-c:v", "libx264", "-pix_fmt", "yuv420p",

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% 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 3)May include surrounding context.

md
---
name: "cc-video-creation"
description: "CC can create professional FactSage videos itself — free, no API, without asking Jarvis. Pipeline: story → animation (Manim/Ken Burns/Kling AI) → upload."
metadata: {"clawbot":{"requires":{"python3":true},"permissions":{"exec":["python3"],"files":["projects/factsage/"],"notes":"Local video production (Manim/Ken Burns). Kling AI is a web UI (manual). No API keys, no paid calls."}}}
---

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% 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 9)May include surrounding context.

md
---
name: "cc-video-creation"
description: "CC can create professional FactSage videos itself — free, no API, without asking Jarvis. Pipeline: story → animation (Manim/Ken Burns/Kling AI) → upload."
metadata: {"clawbot":{"requires":{"python3":true},"permissions":{"exec":["python3"],"files":["projects/factsage/"],"notes":"Local video production (Manim/Ken Burns). Kling AI is a web UI (manual). No API keys, no paid calls."}}}
---

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The documentation exposes a specific third-party account email and instructs users to log into that account without any credential-handling safeguards. This can encourage account sharing, accidental unauthorized access, phishing risk, and leakage of organization-linked identity information.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
74% confidence
Finding

The prompt examples and surrounding instructions are written as fixed English-language generation inputs with no indication that the user may choose another language or locale. Under the stated policy, forcing a specific language without opt-in is a natural-language policy concern unless the constraint is documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The module docstring and command guidance are written entirely in Danish, and the script continues this with Danish user-facing status messages throughout the file. This creates a language-policy issue because the skill imposes a specific locale on users without opt-in, fallback, or a documented regional requirement.

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 · cc_animator.py (reported line 228)May include surrounding context.

python
"-vf", "format=yuv420p",
        str(video_path)
    ]
    result = subprocess.run(cmd, capture_output=True, text=True)
    
    if result.returncode != 0:
        print(f"    ❌ ffmpeg fejl: {result.stderr[-200:]}")

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The header text explicitly emphasizes no external tooling/API and documents commands like '--scene' and '--video'. In reality, the code shells out to ffmpeg via subprocess and the main block does not parse or honor those documented flags, so the documentation actively misstates how the script operates.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The module docstring and user-facing CLI/status text are written in Danish, which imposes a specific language on users without opt-in or an alternative. This is a natural-language locale constraint visible in instructions and runtime output rather than a region-specific requirement that is explicitly justified.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest describes a pipeline ending in upload and mentions broader animation options, but this file only generates PNG frames and invokes ffmpeg to assemble a local MP4. There is no code here for uploading or remote publication, so the implemented behavior is materially narrower than the stated skill description.

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 · cc_illustrator.py (reported line 611)May include surrounding context.

python
def render_video(name="cc-illustrated-09", fps=30):
    print("\n🎞 Renderer video...")
    output = OUTPUT / f"{name}.mp4"
    result = subprocess.run([
        "ffmpeg", "-y", "-framerate", str(fps),
        "-i", str(FRAMES / "frame_%05d.png"),
        "-c:v", "libx264", "-pix_fmt", "yuv420p",

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The user-facing comments and prompts are written exclusively in Danish, including usage instructions and operational prompts. This imposes a specific language on users without opt-in or an alternative, which matches the language/locale policy violation criteria.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The module docstring explicitly states that professional animation is added 'uden AI modeller' and lists local requirements as PIL, numpy, ffmpeg. Later, the implementation imports edge_tts and generates voiceover audio, which directly contradicts the documented claim that the workflow does not use AI-model-based functionality.

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/produce_animated.py (reported line 151)May include surrounding context.

python
'-pix_fmt', 'yuv420p',
            output_path
        ]
        subprocess.run(cmd, capture_output=True, check=True)
    else:
        # Brug PIL til at generere ÉN frame som baggrund
        # og brug ffmpeg zoompan til at animere den

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/produce_animated.py (reported line 183)May include surrounding context.

python
'-pix_fmt', 'yuv420p',
            output_path
        ]
        subprocess.run(cmd, capture_output=True, check=True)
    else:
        # Brug PIL til at generere ÉN frame som baggrund
        # og brug ffmpeg zoompan til at animere den

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/produce_animated.py (reported line 293)May include surrounding context.

python
'-pix_fmt', 'yuv420p',
            output_path
        ]
        subprocess.run(cmd, capture_output=True, check=True)
    else:
        # Brug PIL til at generere ÉN frame som baggrund
        # og brug ffmpeg zoompan til at animere den

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The manifest says the skill can create videos 'free, no API' and the module docstring emphasizes animation 'uden AI modeller' (without AI models). In practice, the code imports edge_tts and uses Communicate.save() to synthesize audio, which is an AI-driven text-to-speech capability and not part of the stated story→animation→upload pipeline.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The function signature defaults text-to-speech generation to "en-US-JennyNeural", which imposes a specific language and locale choice in the skill's behavior. The file does not provide a user opt-in, alternative locale selection path, or justification for why US English is required.

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/produce_animated.py (reported line 365)May include surrounding context.

python
asyncio.run(generate_audio(vo_text, audio_path, voice))
    
    # Get audio duration
    result = subprocess.run(
        ['ffprobe', '-v', 'error', '-show_entries', 'format=duration',
         '-of', 'default=noprint_wrappers=1:nokey=1', audio_path],
        capture_output=True, text=True

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
82% confidence
Finding

The concat demuxer is invoked with -safe 0, and the concat manifest is written using file paths that may contain special characters such as single quotes. While this is not shell injection, disabling ffmpeg's safety checks can allow unintended path handling or inclusion of attacker-influenced files if segment paths or working directories become untrusted, increasing the attack surface around local file access and parser behavior.

Content

Scanner excerpt · scripts/produce_animated.py (reported line 409)May include surrounding context.

python
for seg in segment_files:
                f.write(f"file '{seg}'\n")
        
        subprocess.run([
            'ffmpeg', '-y', '-f', 'concat', '-safe', '0',
            '-i', concat_file, '-c', 'copy', concat_video
        ], capture_output=True, check=True)