Back to skill

Security audit

图片与视频资料标注

Security checks for vulnerabilities and agentic risk

Overview

The skill coherently prepares local media metadata, with disclosed file overwrites and an optional aimeGeo handoff.

Use a separate output directory for each batch to avoid overwriting earlier reports. Expect Chinese report headings and filename suggestions. The optional aimeGeo continuation should use only the accounts and workflows you choose.

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
  • 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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (12)

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: 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.

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/self_test.py (reported line 204)May include surrounding context.

python
data["assets"][0]["duration_seconds"] = value
                    source.write_text(json.dumps(data, ensure_ascii=False), encoding="utf-8")
                    out = temp / str(index)
                    done = subprocess.run([sys.executable, str(ROOT / "scripts/build_metadata.py"),
                                           "--input", str(source), "--output-dir", str(out)],
                                          capture_output=True, text=True)
                    self.assertEqual((done.returncode, done.stderr), (0, ""))

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/self_test.py (reported line 260)May include surrounding context.

python
for raw in (audio()["assets"][0], sample()["assets"][3]):
                raw["upload_date"] = "2026-10-18T10:00:00+08:59"
                source.write_text(json.dumps({"assets": [raw]}), encoding="utf-8")
                success = subprocess.run(cli + ["--output-dir", str(output_dir)],
                                         capture_output=True, text=True)
                self.assertEqual(success.returncode, 0, success.stderr)
                asset = read_document(output_dir / "媒体元信息.json")["assets"][0]

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/self_test.py (reported line 272)May include surrounding context.

python
source.write_text(json.dumps({"assets": [raw]}), encoding="utf-8")
                    for target in (output_dir, rejected_dir):
                        with self.subTest(kind=raw["type"], offset=offset, target=target.name):
                            failure = subprocess.run(cli + ["--output-dir", str(target)],
                                                     capture_output=True, text=True)
                            self.assertEqual(failure.returncode, 2, failure.stderr)
                            self.assertIn("assets[0].upload_date:", failure.stderr)

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/self_test.py (reported line 338)May include surrounding context.

python
def test_cli_success_and_invalid_input_does_not_replace_outputs(self):
        with tempfile.TemporaryDirectory() as temp:
            cli = [sys.executable, str(ROOT / "scripts/build_metadata.py")]
            run = subprocess.run(cli + ["--input", str(ROOT / "examples/success/input.json"),
                                        "--output-dir", temp], capture_output=True, text=True)
            self.assertEqual(run.returncode, 0, run.stderr)
            output = Path(temp) / "媒体元信息.json"

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/self_test.py (reported line 347)May include surrounding context.

python
self.assertEqual((Path(temp) / "图注与字幕交接清单.md").read_text(encoding="utf-8"), handoff(expected))
            self.assertEqual(json.loads(run.stdout), expected["summary"])
            before = output.read_bytes()
            failure = subprocess.run(cli + ["--input", str(ROOT / "examples/failure/unseen-description.json"),
                                            "--output-dir", temp], capture_output=True, text=True)
            self.assertEqual(failure.returncode, 2)
            self.assertIn("unavailable evidence cannot support", failure.stderr)

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/self_test.py (reported line 359)May include surrounding context.

python
for content in ('{"assets":', '{"assets": [], "assets": []}', '{"assets": NaN}'):
                with self.subTest(content=content):
                    source.write_text(content, encoding="utf-8")
                    run = subprocess.run([sys.executable, str(ROOT / "scripts/build_metadata.py"),
                                          "--input", str(source), "--output-dir", str(Path(temp) / "out")],
                                         capture_output=True, text=True)
                    self.assertEqual(run.returncode, 2)

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 (reported line 12)May include surrounding context.

text
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

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 · NOTICE.md (reported line 28)May include surrounding context.

md
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

The skill prescribes Chinese filenames without offering a language choice.

Low
Category
Output Handling
Confidence
92% confidence
Finding

The image deliverables explicitly require Chinese safe filenames, and line 3 likewise specifies Chinese filename suggestions. This imposes a particular output language without user opt-in or a documented locale-specific justification; merely writing the README in Chinese does not establish that choice.

Content

No source excerpt is available for this finding.

The workflow requires Chinese filenames without offering a language choice.

Low
Category
Output Handling
Confidence
99% confidence
Finding

Step 3 explicitly instructs the agent to suggest short Chinese filenames (「建议简短中文文件名」). This imposes a specific output language without user opt-in or a documented locale-specific justification.

Content

No source excerpt is available for this finding.

The generated handoff report is fixed to Chinese without a language choice or opt-in.

Low
Category
Output Handling
Confidence
91% confidence
Finding

The handoff renderer hard-codes Chinese headings, field labels, and explanatory notes regardless of the user's language. The file provides no locale selection or documented regional requirement justifying this fixed output language.

Content

No source excerpt is available for this finding.