Back to skill

Security audit

Social Downloader

Security checks for vulnerabilities and agentic risk

Overview

The skill does what it claims, but its transcription mode can automatically upload extracted audio to OpenAI when an API key is present without clearly telling the user or requiring a separate opt-in.

Review before installing. Use it only for URLs you are authorized to download, run it in a private output directory, and avoid running the transcription script with OPENAI_API_KEY in the environment unless you intend to send the video's audio to OpenAI and accept any privacy and billing implications.

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)

other

Warning
Location
scripts/download_transcribe.py:15
Finding

Automatic Disclosure of Extracted Audio to an External Transcription Service

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/download_transcribe.py:7
Finding

Predictable Output Paths Allow Local Symlink-Based File Overwrite

Content
View full analysis
" --out /tmp/social-skill-source ``` ### Technical Analysis The script accepts or creates an output directory without verifying its ownership, permissions, or whether existing output entries are symbolic links. It then writes predictable filenames including `audio.wav`, `transcript.txt`, and `result.json`. The `ffmpeg` invocation uses `-y`, which enables unconditional overwrite of the destination. If `audio.wav` is an attacker-created symbolic link, the destination followed by `ffmpeg` may be overwritten with the privileges of the account running the skill. Python's `Path.write_text()` normally follows symbolic links as well, exposing `transcript.txt` and `result.json` to the same class of attack. The documented use of `/tmp/social-s ...[truncated 2402 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (10)

Tainted flow: 'req' from os.environ.get (line 30, credential/environment) → urllib.request.urlopen (network output)

Critical
Category
Data Flow
Confidence
90% confidence
Finding

Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.

Content

Scanner excerpt · scripts/download_transcribe.py (reported line 31)May include surrounding context.

python
part('file', audio.read_bytes(), audio.name, 'audio/wav')
    body.append(f'--{boundary}--\r\n'.encode())
    req = urllib.request.Request('https://api.openai.com/v1/audio/transcriptions', data=b''.join(body), headers={'Authorization': f'Bearer {key}', 'Content-Type': f'multipart/form-data; boundary={boundary}'})
    with urllib.request.urlopen(req, timeout=180) as r:
        data = json.loads(r.read().decode())
    (out / 'transcript.txt').write_text(data.get('text',''))
    result['transcript'] = str(out / 'transcript.txt')

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill invokes shell scripts and Python with network, filesystem, and likely environment access, but it declares no explicit tool scope or permissions boundary. In an agent system, this increases the chance the skill is auto-invoked with broader capabilities than intended, enabling unreviewed downloads, local file writes, and shell execution against user-supplied URLs.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The description is broad enough to match many common requests to download, save, send, or transcribe social media content, which can cause the agent to trigger this skill in situations where the user did not clearly consent to local downloading or rights-sensitive processing. Overbroad triggers are dangerous because they expand the operational surface for shell/network actions and increase the chance of downloading untrusted or privacy-sensitive content automatically.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The activation guidance 'Use this for social video links' is ambiguous and does not define supported domains, content restrictions, or required consent steps. In context, that ambiguity matters because the skill performs network retrieval and local storage, so an imprecise trigger can lead to misuse on malicious URLs, non-social links, or content involving legal and privacy risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill tells the agent to download third-party media to local storage and possibly transcribe it, but it does not warn users that this creates local copies and may handle copyrighted, private, or otherwise sensitive material. This omission can undermine informed consent and lead to unsafe handling of personal data or policy-sensitive content, especially when combined with automated download and transcription workflows.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The script invokes yt-dlp to fetch content from a user-supplied URL, which performs a network operation and writes media files into the output directory. There is no confirmation prompt, user-facing print/log message, or inline comment/docstring disclosing that the script will contact a remote service and save files locally.

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/download_transcribe.py (reported line 10)May include surrounding context.

python
args = parser.parse_args()
out = pathlib.Path(args.out); out.mkdir(parents=True, exist_ok=True)
video_tpl = str(out / 'video.%(ext)s')
subprocess.run(['yt-dlp', '-f', 'bv*+ba/best', '--merge-output-format', 'mp4', '-o', video_tpl, args.url], check=True)
mp4 = next(out.glob('video.*'))
audio = out / 'audio.wav'
subprocess.run(['ffmpeg', '-y', '-i', str(mp4), '-vn', '-ac', '1', '-ar', '16000', str(audio)], check=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 · scripts/download_transcribe.py (reported line 13)May include surrounding context.

python
subprocess.run(['yt-dlp', '-f', 'bv*+ba/best', '--merge-output-format', 'mp4', '-o', video_tpl, args.url], check=True)
mp4 = next(out.glob('video.*'))
audio = out / 'audio.wav'
subprocess.run(['ffmpeg', '-y', '-i', str(mp4), '-vn', '-ac', '1', '-ar', '16000', str(audio)], check=True)
result = {'video': str(mp4), 'audio': str(audio)}
key = os.environ.get('OPENAI_API_KEY')
if key:

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

This code performs external transmission of downloaded audio content to a third-party API. In isolation that can be legitimate, but within this skill the transmission is implicit and coupled to the mere presence of a credential, making it easy for users to trigger off-device data sharing without realizing it.

Content

Scanner excerpt · scripts/download_transcribe.py (reported line 30)May include surrounding context.

python
part('response_format', 'json')
    part('file', audio.read_bytes(), audio.name, 'audio/wav')
    body.append(f'--{boundary}--\r\n'.encode())
    req = urllib.request.Request('https://api.openai.com/v1/audio/transcriptions', data=b''.join(body), headers={'Authorization': f'Bearer {key}', 'Content-Type': f'multipart/form-data; boundary={boundary}'})
    with urllib.request.urlopen(req, timeout=180) as r:
        data = json.loads(r.read().decode())
    (out / 'transcript.txt').write_text(data.get('text',''))

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

If OPENAI_API_KEY is present, the script automatically uploads the extracted audio to an external transcription service without any explicit user notice, consent, or opt-in at runtime. In a social-video downloader context, downloaded media can contain personal, copyrighted, or sensitive voice content, so silent transmission materially increases privacy and compliance risk.

Content

No source excerpt is available for this finding.