Back to skill

Security audit

本地 Vlog 制作 / Local Vlog Production

Security checks for vulnerabilities and agentic risk

Overview

The verdict is suspicious because scanners flagged a privacy concern about sending story-derived prompts to image-generation services and a confirmed risk that media tools can run without time limits; the evidence is incomplete because Skillspector reported partial analysis.

Consider the privacy implications before using image generation with personal stories, and check whether the service is local or external and what prompt text may be shared. Also consider the reported risk of media processing running without a time limit. Dependency versions are unpinned, and Skillspector's analysis is marked partial, so the available evidence does not establish that the artifact is safe.

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
  • 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
Findings (23)

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · tests/test_workspace.py (reported line 71)May include surrounding context.

python
checks = []
    logs = case / "test_logs"
    logs.mkdir()
    env = dict(os.environ, PYTHONDONTWRITEBYTECODE="1")
    def run_case(label, script, arguments, expected=0):
        result = subprocess.run([sys.executable, str(BUNDLE / "tools" / script), "--root", str(case), *arguments], capture_output=True, encoding="utf-8", errors="replace", timeout=180, env=env, cwd=case)
        filename = f"{len(checks)+1:02d}.log"

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 · tools/make_demo.py (reported line 78)May include surrounding context.

python
wav=output/'demo_audio.wav'
    with wave.open(str(wav),'wb') as out:
        out.setnchannels(1);out.setsampwidth(2);out.setframerate(sr);out.writeframes(pcm.tobytes())
    result=subprocess.run([ffmpeg,'-hide_banner','-v','error','-nostdin','-n','-i',str(wav),
        '-c:a','aac','-b:a','128k',str(output/'demo_audio.m4a')],capture_output=True)
    if result.returncode: raise RuntimeError(result.stderr.decode('utf-8','replace'))
    (output/'README.txt').write_text('全部为程序生成的测试图形和低电平合成音,不是真实照片、用户语音或配乐。\n'

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, file_read, file_write, network, shell.

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.

The workflow instructs image generation without warning that story text or personal details in prompts may be shared with the generation service.

Medium
Category
Output Handling
Confidence
83% confidence
Finding

The workflow directs the agent to send story-derived text to an available image-generation tool without requiring disclosure or authorization when that tool uses an external service. In a personal Vlog workflow, prompts can contain names, locations, or private experiences, so even text-only generation can cross a privacy boundary; the broader skill's data-minimization instructions reduce but do not eliminate this risk.

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 · tests/selftest.py (reported line 46)May include surrounding context.

python
report={'status':'running','synthetic_only':True,'checks':[],
            'continuous_playback':False,'actual_listening':False,'real_source_media_used':False}
    def invoke(script,*extra):
        result=subprocess.run([sys.executable,'-X','utf8',str(relocated/script),*map(str,extra)],cwd=relocated,capture_output=True)
        if result.returncode:
            raise RuntimeError(result.stderr.decode('utf-8','replace')[-5000:]+'\n'+result.stdout.decode('utf-8','replace')[-3000:])
        return 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 · tests/selftest.py (reported line 83)May include surrounding context.

python
assert not qa['continuous_playback'] and not qa['actual_listening']
        passed('full_render_decode_and_expected_duration',frames=240,duration_seconds=8)
        def packets(path):
            data=subprocess.check_output([fp,'-v','error','-select_streams','a:0','-show_packets','-show_data_hash','sha256',
                '-show_entries','packet=pts_time,dts_time,duration_time,data_hash','-of','json',str(path)])
            return json.loads(data)['packets']
        before=packets(relocated/'examples/media/demo_audio.m4a')

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 · tests/selftest.py (reported line 91)May include surrounding context.

python
assert before==after, 'Accepted master AAC content or timestamps changed'
        passed('aac_packet_content_and_timestamps_preserved',packet_count=len(after))
        def frame(seconds):
            data=subprocess.check_output([ff,'-v','error','-ss',str(seconds),'-i',str(relocated/video),'-frames:v','1','-f','rawvideo','-pix_fmt','rgb24','-'])
            return np.frombuffer(data,dtype=np.uint8).reshape((640,360,3))
        a,b=frame(1.5),frame(2.3)
        movement=float(np.abs(a[505:550].astype(float)-b[505:550].astype(float)).mean())

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 · tests/test_qa.py (reported line 32)May include surrounding context.

python
(cls.root / "workspace_config.json").write_text(json.dumps({"executables": {"ffmpeg": FF, "ffprobe": FP}}), encoding="utf-8")

        def create(*args):
            subprocess.run([FF, "-hide_banner", "-loglevel", "error", "-nostdin", "-n", *args],
                           check=True, cwd=cls.root, timeout=30)
        create("-f", "lavfi", "-i", "color=white:s=320x180:r=24:d=7",
               "-f", "lavfi", "-i", "sine=f=440:r=48000:d=7", "-f", "lavfi", "-i", "sine=f=880:r=48000:d=7",

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 · tests/test_workspace.py (reported line 63)May include surrounding context.

python
Image.new("RGB", (120, 80), "#76b4c9").save(theme / "IMG_20260102_030405.jpg", exif=exif)
    Image.new("RGB", (80, 120), "#f4cf68").save(theme / "photo.png")
    write_text_new(theme / "clip.txt", "Synthetic companion; not real footage.\n")
    creation = subprocess.run([ffmpeg, "-hide_banner", "-v", "error", "-nostdin", "-f", "lavfi", "-i", "testsrc2=size=120x80:rate=10", "-t", "2", "-an", "-c:v", "libx264", "-pix_fmt", "yuv420p", "-n", str(theme / "clip.mp4")], capture_output=True, encoding="utf-8", errors="replace", timeout=60)
    if creation.returncode:
        raise RuntimeError("合成视频创建失败:" + creation.stderr[-2000:])
    write_text_new(case / "brief.txt", "Synthetic test only; no private material.\n")

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 · tests/test_workspace.py (reported line 73)May include surrounding context.

python
logs.mkdir()
    env = dict(os.environ, PYTHONDONTWRITEBYTECODE="1")
    def run_case(label, script, arguments, expected=0):
        result = subprocess.run([sys.executable, str(BUNDLE / "tools" / script), "--root", str(case), *arguments], capture_output=True, encoding="utf-8", errors="replace", timeout=180, env=env, cwd=case)
        filename = f"{len(checks)+1:02d}.log"
        write_text_new(logs / filename, result.stdout + "\nSTDERR:\n" + result.stderr)
        checks.append({"name": label, "expected_exit": expected, "actual_exit": result.returncode, "passed": result.returncode == expected, "log": "test_logs/" + filename})

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 · tools/make_demo.py (reported line 54)May include surrounding context.

python
'-colorspace','bt709','-color_range','tv',
        '-x264-params','colorprim=bt709:transfer=bt709:colormatrix=bt709',str(output/'demo_video.mp4')]
    with (output/'demo_video.encoder.log').open('wb') as error:
        process=subprocess.Popen(command,stdin=subprocess.PIPE,stderr=error)
        try:
            small = ImageFont.truetype(str(font_path),22)
            for i in range(fps*seconds):

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 · tools/make_demo.py (reported line 78)May include surrounding context.

python
wav=output/'demo_audio.wav'
    with wave.open(str(wav),'wb') as out:
        out.setnchannels(1);out.setsampwidth(2);out.setframerate(sr);out.writeframes(pcm.tobytes())
    result=subprocess.run([ffmpeg,'-hide_banner','-v','error','-nostdin','-n','-i',str(wav),
        '-c:a','aac','-b:a','128k',str(output/'demo_audio.m4a')],capture_output=True)
    if result.returncode: raise RuntimeError(result.stderr.decode('utf-8','replace'))
    (output/'README.txt').write_text('全部为程序生成的测试图形和低电平合成音,不是真实照片、用户语音或配乐。\n'

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 · tools/qa.py (reported line 27)May include surrounding context.

python
def probe_streams(executable, source, timeout):
    fields = "format=duration:stream=" + ",".join(FIELDS) + ":stream_disposition=" + ",".join(DISPOSITIONS)
    with tempfile.TemporaryFile() as errors:
        probe = subprocess.run([executable, "-v", "error", "-show_entries", fields, "-of", "json", source],
                               stdin=subprocess.DEVNULL, stdout=subprocess.PIPE, stderr=errors, timeout=timeout)
        if probe.returncode:
            errors.seek(max(0, errors.tell() - 4000))

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 · tools/qa.py (reported line 104)May include surrounding context.

python
finished.set()

    stopped = False
    with subprocess.Popen(command, stdin=subprocess.DEVNULL, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
                          text=True, encoding="utf-8", errors="replace") as proc:
        worker = threading.Thread(target=drain, args=(proc.stdout,), daemon=True)
        worker.start()

Unbounded Resource Access

Medium
Category
Excessive Agency
Confidence
95% confidence
Finding

The helper defaults to no timeout, and its shown callers do not supply one, allowing malformed or expensive media to stall probing, muxing, hashing, or validation indefinitely. It also captures stdout and stderr entirely in memory, including potentially large packet-evidence JSON, so hostile inputs can exhaust memory and deny service to the local production workflow.

Content

Scanner excerpt · tools/render_timeline.py (reported line 92)May include surrounding context.

python
raise TimelineError(f"Unsupported {label} fields: {', '.join(sorted(unknown))}")


def run(command, timeout=None):
    result = subprocess.run([str(item) for item in command], stdout=subprocess.PIPE,
                            stderr=subprocess.PIPE, timeout=timeout)
    if result.returncode:

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 · tools/render_timeline.py (reported line 93)May include surrounding context.

python
def run(command, timeout=None):
    result = subprocess.run([str(item) for item in command], stdout=subprocess.PIPE,
                            stderr=subprocess.PIPE, timeout=timeout)
    if result.returncode:
        # Commands can contain local paths; public manifests store only a short diagnostic.

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 · tools/render_timeline.py (reported line 382)May include surrounding context.

python
"-x264-params", "colorprim=bt709:transfer=bt709:colormatrix=bt709",
                       "-color_primaries", "bt709", "-color_trc", "bt709", "-colorspace", "bt709", "-color_range", "tv",
                       "-movflags", "+faststart", destination]
            encoder = subprocess.Popen([str(value) for value in command], stdin=subprocess.PIPE, stderr=log)
            for index, clip in enumerate(plan["clips"]):
                decoder = None
                with open(tempdir / f"decode-{index}.log", "wb") as decode_log:

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 · tools/render_timeline.py (reported line 391)May include surrounding context.

python
command = [ffmpeg, "-hide_banner", "-loglevel", "error", "-ss", clip.get("source_in", 0),
                                       "-i", resolve(clip["source"]), "-map", "0:v:0", "-an", "-sn", "-dn",
                                       "-vf", video_filter(clip, plan), "-frames:v", clip["frames"], "-f", "rawvideo", "-pix_fmt", "rgb24", "pipe:1"]
                            decoder = subprocess.Popen([str(value) for value in command], stdout=subprocess.PIPE, stderr=decode_log)
                        elif clip["type"] == "image":
                            still = fit_image(load_image(resolve(clip["source"])), (width, height), clip.get("fit", "cover"),
                                              clip.get("crop_center", [0.5, 0.5]), plan["background"])

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 · tools/workspace.py (reported line 138)May include surrounding context.

python
def run_command(command, timeout=60):
    completed = subprocess.run(command, capture_output=True, encoding="utf-8", errors="replace", timeout=timeout)
    return completed.returncode, completed.stdout, completed.stderr

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · LICENSE.txt (reported line 7)May include surrounding context.

text
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Unverifiable Dependency: Pillow has 214 known advisory(ies) (CVE-2026-112390 (Pillow: Infinite loop in MIC n_frames when the embedded TIFF); CVE-2026-96638 (Pillow: BMP RLE decoder memory-amplification DoS via unbounded delta escape / ab); CVE-2026-112391 (Pillow: PdfParser can recurse without limit in get_value() and linearize_page_tr) +211 more), but the manifest does not pin a version, so it is unknown whether the installed release is affected

Low
Category
Supply Chain
Confidence
40% confidence
Finding

OSV returned advisories for this package, but its resolved version is unknown. A package-name lookup does not establish whether the installed release is affected by those advisories.

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
40% confidence
Finding

OSV returned advisories for this package, but its resolved version is unknown. A package-name lookup does not establish whether the installed release is affected by those advisories.

Content

No source excerpt is available for this finding.

Dynamic attribute access via getattr()

Low
Category
Dangerous Code Execution
Confidence
50% confidence
Finding

Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.

Content

Scanner excerpt · tests/test_workspace.py (reported line 29)May include surrounding context.

python
args = parser.parse_args()
    workspace = Workspace(args.root)
    def executable(name):
        value = getattr(args, name)
        if value:
            candidate = shutil.which(value)
            if not candidate and Path(value).is_file():