Back to skill

Security audit

Bravado

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent video-production helper that uses expected web lookup, local rendering, and file output without evidence of hidden persistence, credential access, or data theft.

Install if you are comfortable with a video-generation skill that may fetch public project URLs, query GitHub metadata for repository links, download verified fonts, run local Python/ffmpeg commands, and write editable render artifacts. Avoid giving it private resumes, portfolios, or repositories unless you intend those materials to be processed for the film.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (14)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The description promises a broad creative production workflow for generating polished product motion films from external project materials. In contrast, the code chunk is narrowly focused on validating the structure and contents of a JSON brief against a fixed schema used by a starter renderer. It performs no external content inspection, no media generation, no storyboard creation, no sound design, and no delivery of video artifacts. This is a material mismatch in primary purpose and capabilities, not merely an implementation detail.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The code chunk does not implement the declared core functionality of creating a polished product motion film or any of its described workflow steps. Instead, it only downloads and verifies specific font files for a renderer. While font setup could be a supporting implementation detail within a larger video-rendering system, this isolated code chunk’s actual purpose is materially narrower and different from the declared skill description. It also performs network access to retrieve fonts, which is not reflected in the declared permissions or description. Therefore this is a clear description-behavior mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared description presents a full creative production workflow for making branded motion films from user-provided project materials. The actual code does none of that. It only validates an already-produced MP4 export against technical delivery requirements and decodes the file to catch corruption. While export verification could be a supporting step in a video-production pipeline, this code chunk’s primary purpose is technical QA, not creation of a motion film. That is a materially different primary behavior from the declared purpose.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared description promises a creative production skill for making branded motion videos from user-provided project materials. The actual code does not analyze any external project input, create scripts/storyboards, render media, or produce output files like MP4s. Its sole behavior is local filesystem setup: validating a destination directory and copying starter assets into it. That is a materially different primary purpose, so this is a clear description-behavior mismatch.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding

The skill instructs the agent to read/write files, access external URLs, and run shell commands, but the manifest does not declare any tool scope or allowed-tools restrictions. That creates unnecessary ambiguity about what capabilities may be exercised at runtime and weakens least-privilege controls, especially because the workflow includes network inspection and local script execution.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest uses broad activation language like 'use when the user asks for a Bravado video, launch intro, explainer' that can match many ordinary video-related requests. Over-broad triggering can cause the agent to invoke this skill in contexts the user did not intend, leading to unnecessary network access, file operations, or shell execution described elsewhere in the workflow.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill enables implicit invocation without defining trigger constraints or exclusion conditions, so it may activate in response to loosely related user requests. Because this skill fetches and inspects external project links and then performs multi-step content generation, unintended activation could cause unnecessary network access, processing of untrusted content, or user surprise around data handling and side effects.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · assets/starter/inspect_source.py (reported line 48)May include surrounding context.

python
parts=[p for p in parsed.path.split('/') if p]
        if len(parts)<2:raise ValueError('Provide a GitHub repository URL, not just a user profile')
        owner,repo=parts[:2];output['source_type']='github_repository'
        api=f'https://api.github.com/repos/{owner}/{repo}'
        body,final=get(api);data=json.loads(body)
        output.update({'redirected_url':final,'name':data.get('name'),'description':data.get('description'),
                       'homepage':data.get('homepage'),'default_branch':data.get('default_branch'),

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · assets/starter/inspect_source.py (reported line 54)May include surrounding context.

python
parts=[p for p in parsed.path.split('/') if p]
        if len(parts)<2:raise ValueError('Provide a GitHub repository URL, not just a user profile')
        owner,repo=parts[:2];output['source_type']='github_repository'
        api=f'https://api.github.com/repos/{owner}/{repo}'
        body,final=get(api);data=json.loads(body)
        output.update({'redirected_url':final,'name':data.get('name'),'description':data.get('description'),
                       'homepage':data.get('homepage'),'default_branch':data.get('default_branch'),

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 · assets/starter/render.py (reported line 310)May include surrounding context.

python
with wave.open(str(OUT/'sound-design.wav'),'wb') as w:
        w.setnchannels(2);w.setsampwidth(2);w.setframerate(sr);w.writeframes((np.clip(a,-1,1)*32767).astype('<i2').tobytes())
    normalized=OUT/'sound-normalized.wav'
    subprocess.run(['ffmpeg','-y','-v','error','-i',str(OUT/'sound-design.wav'),'-af','loudnorm=I=-18:TP=-1.5:LRA=9','-ar','48000',str(normalized)],check=True)
    normalized.replace(OUT/'sound-design.wav')

def main():

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 · assets/starter/render.py (reported line 343)May include surrounding context.

python
name=''.join(ch.lower() if ch.isalnum() else '-' for ch in b('brand','name')).strip('-')
    output=OUT/f'{name}-bravado.mp4'
    cmd=['ffmpeg','-y','-v','error','-f','rawvideo','-pix_fmt','rgb24','-s',f'{W}x{H}','-r',str(FPS),'-i','-', '-i',str(OUT/'sound-design.wav'),'-c:v','libx264','-preset','fast','-crf','18','-pix_fmt','yuv420p','-c:a','aac','-b:a','192k','-movflags','+faststart','-t',str(DURATION),str(output)]
    p=subprocess.Popen(cmd,stdin=subprocess.PIPE)
    for i in range(FPS*DURATION):
        p.stdin.write(frame(i/FPS).tobytes())
        if i%150==0:print(f'Rendered {i}/{FPS*DURATION} frames',flush=True)

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 · assets/starter/verify_export.py (reported line 95)May include surrounding context.

python
def _run(command: list[str], stage: str) -> str:
    try:
        result = subprocess.run(command, capture_output=True, text=True, check=False)
    except FileNotFoundError as exc:
        raise ExportVerificationError(f"{stage}: {command[0]} is not installed or not on PATH") from exc
    except OSError as exc:

Unverifiable Dependency: Pillow has 16 known advisory(ies) (CVE-2016-2533 (Pillow buffer overflow in ImagingPcdDecode); CVE-2023-50447 (Arbitrary Code Execution in Pillow); CVE-2021-27922 (Pillow Uncontrolled Resource Consumption) +13 more), but the manifest does not pin a version, so it is unknown whether the installed release is affected

Low
Category
Supply Chain
Confidence
91% confidence
Finding

The dependency is only range-pinned (>=10,<13), not locked to an exact vetted version, so the actual installed Pillow release may vary across environments and could include a vulnerable build within the allowed range. Because Pillow processes image data and this skill is for rendering motion/video assets, any vulnerable image parsing path is more relevant than in a non-media skill.

Content

No source excerpt is available for this finding.

Unverifiable Dependency: numpy has 16 known advisory(ies) (CVE-2014-1859 (Numpy arbitrary file write via symlink attack); CVE-2021-41495 (NumPy NULL Pointer Dereference); CVE-2021-33430 (NumPy Buffer Overflow (Disputed)) +13 more), but the manifest does not pin a version, so it is unknown whether the installed release is affected

Low
Category
Supply Chain
Confidence
85% confidence
Finding

NumPy is specified as a version range (>=1.26,<3) rather than an exact approved version, so different installations may resolve to different releases, making it impossible to verify whether a vulnerable version is used. The risk is somewhat contextualized by this being a media-generation skill, where NumPy is likely used for local numerical processing rather than direct untrusted parsing, but supply-chain uncertainty remains.

Content

No source excerpt is available for this finding.